{"id":3595,"date":"2025-05-03T14:34:40","date_gmt":"2025-05-03T11:34:40","guid":{"rendered":"https:\/\/www.dztecs.com\/ai\/?p=3595"},"modified":"2025-06-13T08:50:19","modified_gmt":"2025-06-13T05:50:19","slug":"how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals","status":"publish","type":"post","link":"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/","title":{"rendered":"\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0646\u0645\u0627\u0630\u062c DeepSeek-R1 \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0639\u0644\u0649 GPQA \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 Ollama \u0648 simple-evals \u0645\u0646 OpenAI","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"<p class=\"wp-block-paragraph\">\u0623\u062d\u062f\u062b \u0625\u0637\u0644\u0627\u0642 \u0644\u0646\u0645\u0648\u0630\u062c <a href=\"https:\/\/api-docs.deepseek.com\/news\/news250120\" target=\"_blank\" rel=\"noreferrer noopener\">DeepSeek-R1<\/a> \u0635\u062f\u0649 \u0648\u0627\u0633\u0639\u064b\u0627 \u0641\u064a \u0645\u062c\u062a\u0645\u0639 \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0627\u0644\u0639\u0627\u0644\u0645\u064a. \u0644\u0642\u062f \u062d\u0642\u0642 \u0627\u062e\u062a\u0631\u0627\u0642\u0627\u062a \u062a\u0636\u0627\u0647\u064a \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0645\u0646 Meta \u0648 OpenAI\u060c \u0648\u062d\u0642\u0642 \u0630\u0644\u0643 \u0641\u064a \u062c\u0632\u0621 \u0635\u063a\u064a\u0631 \u0645\u0646 \u0627\u0644\u0648\u0642\u062a \u0648\u0628\u062a\u0643\u0644\u0641\u0629 \u0623\u0642\u0644 \u0628\u0643\u062b\u064a\u0631.<\/p>\n<p class=\"wp-block-paragraph\">\u0648\u0644\u0643\u0646 \u0628\u0639\u064a\u062f\u064b\u0627 \u0639\u0646 \u0627\u0644\u0639\u0646\u0627\u0648\u064a\u0646 \u0627\u0644\u0631\u0626\u064a\u0633\u064a\u0629 \u0648\u0627\u0644\u0636\u062c\u0629 \u0627\u0644\u0625\u0639\u0644\u0627\u0645\u064a\u0629 \u0639\u0628\u0631 \u0627\u0644\u0625\u0646\u062a\u0631\u0646\u062a\u060c \u0643\u064a\u0641 \u064a\u0645\u0643\u0646\u0646\u0627 \u062a\u0642\u064a\u064a\u0645 \u0642\u062f\u0631\u0627\u062a \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0641\u064a \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0645\u0639\u0627\u064a\u064a\u0631 \u0645\u0639\u062a\u0631\u0641 \u0628\u0647\u0627\u061f \u0647\u0630\u0627 \u0633\u0624\u0627\u0644 \u0645\u0647\u0645 \u0644\u062e\u0628\u0631\u0627\u0621 \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a.<\/p>\n<p class=\"wp-block-paragraph\">\u0648\u0627\u062c\u0647\u0629 \u0627\u0644\u0645\u0633\u062a\u062e\u062f\u0645 \u0627\u0644\u062e\u0627\u0635\u0629 \u0628\u0640 <a href=\"http:\/\/chat.deepseek.com\/\" target=\"_blank\" rel=\"noreferrer noopener\">Deepseek<\/a> \u062a\u062c\u0639\u0644 \u0645\u0646 \u0627\u0644\u0633\u0647\u0644 \u0627\u0633\u062a\u0643\u0634\u0627\u0641 \u0642\u062f\u0631\u0627\u062a\u0647\u0627\u060c \u0648\u0644\u0643\u0646 \u0627\u0633\u062a\u062e\u062f\u0627\u0645\u0647\u0627 \u0628\u0631\u0645\u062c\u064a\u064b\u0627 \u064a\u0648\u0641\u0631 \u0631\u0624\u0649 \u0623\u0639\u0645\u0642 \u0648\u062a\u0643\u0627\u0645\u0644\u064b\u0627 \u0623\u0643\u062b\u0631 \u0633\u0644\u0627\u0633\u0629 \u0641\u064a \u0627\u0644\u062a\u0637\u0628\u064a\u0642\u0627\u062a \u0627\u0644\u0648\u0627\u0642\u0639\u064a\u0629. \u0625\u0646 \u0641\u0647\u0645 \u0643\u064a\u0641\u064a\u0629 \u062a\u0634\u063a\u064a\u0644 \u0647\u0630\u0647 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0645\u062d\u0644\u064a\u064b\u0627 \u064a\u0648\u0641\u0631 \u0623\u064a\u0636\u064b\u0627 \u062a\u062d\u0643\u0645\u064b\u0627 \u0645\u062d\u0633\u0646\u064b\u0627 \u0648\u0648\u0635\u0648\u0644\u0627\u064b \u062f\u0648\u0646 \u0627\u062a\u0635\u0627\u0644 \u0628\u0627\u0644\u0625\u0646\u062a\u0631\u0646\u062a.<\/p>\n<p><img decoding=\"async\" class=\"attachment-post-thumbnail size-post-thumbnail not-transparent wp-post-image aligncenter\" src=\"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2025\/05\/todd-cravens-lwACYK8ScmA-unsplash.jpg\" alt=\"\" width=\"1920\" height=\"1277\" data-has-transparency=\"false\" data-dominant-color=\"6a8fa0\" \/><\/p>\n<p class=\"wp-block-paragraph\">\u0641\u064a \u0647\u0630\u0647 \u0627\u0644\u0645\u0642\u0627\u0644\u0629\u060c \u0633\u0648\u0641 \u0646\u0633\u062a\u0643\u0634\u0641 \u0643\u064a\u0641\u064a\u0629 \u0627\u0633\u062a\u062e\u062f\u0627\u0645 <strong>Ollama<\/strong> \u0648 <strong>simple-evals \u0645\u0646 OpenAI<\/strong> \u0644\u062a\u0642\u064a\u064a\u0645 \u0642\u062f\u0631\u0627\u062a \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0644\u0646\u0645\u0627\u0630\u062c DeepSeek-R1 \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0628\u0646\u0627\u0621\u064b \u0639\u0644\u0649 \u0645\u0639\u064a\u0627\u0631 <strong>GPQA-Diamond<\/strong> \u0627\u0644\u0634\u0647\u064a\u0631. \u064a\u0639\u062a\u0628\u0631 \u0647\u0630\u0627 \u0627\u0644\u0645\u0639\u064a\u0627\u0631 \u0645\u0646 \u0623\u0647\u0645 \u0627\u0644\u0623\u062f\u0648\u0627\u062a \u0644\u062a\u0642\u064a\u064a\u0645 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0641\u064a \u0645\u062c\u0627\u0644 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0645\u0646\u0637\u0642\u064a.<\/p>\n<p class=\"wp-block-paragraph\">\u0625\u0644\u064a\u0643 <a href=\"https:\/\/github.com\/kennethleungty\/DeepSeek-R1-Ollama-Simple-Evals\" target=\"_blank\" rel=\"noreferrer noopener\">\u0631\u0627\u0628\u0637 \u0645\u0633\u062a\u0648\u062f\u0639 GitHub<\/a> \u0627\u0644\u0645\u0635\u0627\u062d\u0628 \u0644\u0647\u0630\u0647 \u0627\u0644\u0645\u0642\u0627\u0644\u0629.<\/p>\n<h2 id=\"heading-one\" class=\"wp-block-heading\">(1) \u0645\u0627 \u0647\u064a \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644\u061f<\/h2>\n<p class=\"wp-block-paragraph\">\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644\u060c \u0645\u062b\u0644 DeepSeek-R1 \u0648\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0633\u0644\u0633\u0644\u0629 o \u0645\u0646 OpenAI (\u0645\u062b\u0644 o1\u060c o3)\u060c \u0647\u064a \u0646\u0645\u0627\u0630\u062c \u0644\u063a\u0648\u064a\u0629 \u0643\u0628\u064a\u0631\u0629 (LLMs) \u062a\u0645 \u062a\u062f\u0631\u064a\u0628\u0647\u0627 \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0627\u0644\u062a\u0639\u0644\u0645 \u0627\u0644\u0645\u0639\u0632\u0632 \u0644\u0623\u062f\u0627\u0621 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644. \u062a\u0639\u062a\u0628\u0631 \u0647\u0630\u0647 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0623\u062f\u0648\u0627\u062a \u0645\u062a\u0637\u0648\u0631\u0629 \u0641\u064a \u0645\u062c\u0627\u0644 \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a\u060c \u062d\u064a\u062b \u062a\u0645\u062b\u0644 \u0642\u0645\u0629 \u0627\u0644\u062a\u0637\u0648\u0631 \u0641\u064a \u0642\u062f\u0631\u0629 \u0627\u0644\u0622\u0644\u0627\u062a \u0639\u0644\u0649 \u0627\u0644\u062a\u0641\u0643\u064a\u0631 \u0627\u0644\u0645\u0646\u0637\u0642\u064a \u0648\u062d\u0644 \u0627\u0644\u0645\u0634\u0643\u0644\u0627\u062a \u0627\u0644\u0645\u0639\u0642\u062f\u0629.<\/p>\n<p class=\"wp-block-paragraph\">\u062a\u062a\u0645\u064a\u0632 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0628\u0627\u0644\u062a\u0641\u0643\u064a\u0631 \u0627\u0644\u0639\u0645\u064a\u0642 \u0642\u0628\u0644 \u0627\u0644\u0625\u062c\u0627\u0628\u0629\u060c \u062d\u064a\u062b \u062a\u0646\u062a\u062c \u0633\u0644\u0633\u0644\u0629 \u0637\u0648\u064a\u0644\u0629 \u0645\u0646 \u0627\u0644\u0623\u0641\u0643\u0627\u0631 \u0627\u0644\u062f\u0627\u062e\u0644\u064a\u0629 \u0642\u0628\u0644 \u0627\u0644\u0627\u0633\u062a\u062c\u0627\u0628\u0629. \u0625\u0646\u0647\u0627 \u062a\u062a\u0641\u0648\u0642 \u0641\u064a \u062d\u0644 \u0627\u0644\u0645\u0634\u0643\u0644\u0627\u062a \u0627\u0644\u0645\u0639\u0642\u062f\u0629\u060c \u0648\u0627\u0644\u0628\u0631\u0645\u062c\u0629\u060c \u0648\u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0639\u0644\u0645\u064a\u060c \u0648\u0627\u0644\u062a\u062e\u0637\u064a\u0637 \u0645\u062a\u0639\u062f\u062f \u0627\u0644\u062e\u0637\u0648\u0627\u062a \u0644\u0633\u064a\u0631 \u0639\u0645\u0644 \u0627\u0644\u0648\u0643\u0644\u0627\u0621. \u0647\u0630\u0647 \u0627\u0644\u0642\u062f\u0631\u0627\u062a \u062a\u062c\u0639\u0644\u0647\u0627 \u0644\u0627 \u063a\u0646\u0649 \u0639\u0646\u0647\u0627 \u0641\u064a \u0645\u062c\u0627\u0644\u0627\u062a \u0645\u062b\u0644 \u062a\u0637\u0648\u064a\u0631 \u0627\u0644\u0628\u0631\u0645\u062c\u064a\u0627\u062a \u0627\u0644\u0645\u062a\u0642\u062f\u0645\u0629\u060c \u0648\u0627\u0644\u0628\u062d\u062b \u0627\u0644\u0639\u0644\u0645\u064a\u060c \u0648\u0623\u062a\u0645\u062a\u0629 \u0627\u0644\u0639\u0645\u0644\u064a\u0627\u062a \u0627\u0644\u0645\u0639\u0642\u062f\u0629.<\/p>\n<h2 id=\"heading-two\" class=\"wp-block-heading\">(2) \u0645\u0627 \u0647\u0648 \u0646\u0645\u0648\u0630\u062c DeepSeek-R1\u061f<\/h2>\n<p class=\"wp-block-paragraph\">DeepSeek-R1 \u0647\u0648 \u0646\u0645\u0648\u0630\u062c \u0644\u063a\u0648\u064a \u0643\u0628\u064a\u0631 \u0645\u0641\u062a\u0648\u062d \u0627\u0644\u0645\u0635\u062f\u0631 (LLM) \u064a\u0639\u062a\u0628\u0631 \u0627\u0644\u0623\u062d\u062f\u062b \u0641\u064a \u0627\u0644\u0645\u062c\u0627\u0644\u060c \u0648\u0645\u0635\u0645\u0645 \u062e\u0635\u064a\u0635\u064b\u0627 \u0644\u0640 <strong>\u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0645\u062a\u0642\u062f\u0645<\/strong>. \u062a\u0645 \u062a\u0642\u062f\u064a\u0645\u0647 \u0641\u064a \u064a\u0646\u0627\u064a\u0631 2025 \u0641\u064a \u0627\u0644\u0648\u0631\u0642\u0629 \u0627\u0644\u0628\u062d\u062b\u064a\u0629 <em>&#8220;<\/em><a href=\"https:\/\/arxiv.org\/abs\/2501.12948\" target=\"_blank\" rel=\"noreferrer noopener\"><em>DeepSeek-R1: \u062a\u062d\u0641\u064a\u0632 \u0642\u062f\u0631\u0629 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0641\u064a \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0648\u064a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631\u0629 \u0639\u0628\u0631 \u0627\u0644\u062a\u0639\u0644\u0645 \u0627\u0644\u0645\u0639\u0632\u0632<\/em><\/a><em>&#8220;<\/em>. \u064a\u0639\u062a\u0628\u0631 DeepSeek-R1 \u0646\u0645\u0648\u0630\u062c\u064b\u0627 \u0631\u0627\u0626\u062f\u064b\u0627 \u0641\u064a \u0645\u062c\u0627\u0644 \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a.<\/p>\n<p class=\"wp-block-paragraph\">\u064a\u0639\u062a\u0645\u062f \u0647\u0630\u0627 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0639\u0644\u0649 \u0645\u0639\u0645\u0627\u0631\u064a\u0629 \u0646\u0645\u0648\u0630\u062c \u0644\u063a\u0648\u064a \u0643\u0628\u064a\u0631 (LLM) \u0628\u0640 671 \u0645\u0644\u064a\u0627\u0631 \u0645\u0639\u0644\u0645\u0629\u060c \u0648\u062a\u0645 \u062a\u062f\u0631\u064a\u0628\u0647 \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0645\u0643\u062b\u0641 \u0644\u0644\u062a\u0639\u0644\u0645 \u0627\u0644\u0645\u0639\u0632\u0632 (RL) \u0628\u0646\u0627\u0621\u064b \u0639\u0644\u0649 \u0627\u0644\u0645\u0633\u0627\u0631 \u0627\u0644\u062a\u0627\u0644\u064a:<\/p>\n<ul class=\"wp-block-list\">\n<li class=\"wp-block-list-item\">\u0645\u0631\u062d\u0644\u062a\u0627\u0646 \u0645\u0646 \u0627\u0644\u062a\u0639\u0632\u064a\u0632 \u062a\u0647\u062f\u0641\u0627\u0646 \u0625\u0644\u0649 \u0627\u0643\u062a\u0634\u0627\u0641 \u0623\u0646\u0645\u0627\u0637 \u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0645\u062d\u0633\u0646\u0629 \u0648\u0627\u0644\u062a\u0648\u0627\u0641\u0642 \u0645\u0639 \u0627\u0644\u062a\u0641\u0636\u064a\u0644\u0627\u062a \u0627\u0644\u0628\u0634\u0631\u064a\u0629.<\/li>\n<li class=\"wp-block-list-item\">\u0645\u0631\u062d\u0644\u062a\u0627\u0646 \u0645\u0646 \u0627\u0644\u0636\u0628\u0637 \u0627\u0644\u062f\u0642\u064a\u0642 \u0627\u0644\u062e\u0627\u0636\u0639 \u0644\u0644\u0625\u0634\u0631\u0627\u0641 \u062a\u0639\u0645\u0644\u0627\u0646 \u0643\u0628\u0630\u0631\u0629 \u0644\u0642\u062f\u0631\u0627\u062a \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0641\u064a \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0648\u063a\u064a\u0631 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644.<\/li>\n<\/ul>\n<p class=\"wp-block-paragraph\">\u0644\u062a\u0648\u0636\u064a\u062d \u0627\u0644\u0623\u0645\u0631\u060c \u0642\u0627\u0645\u062a DeepSeek \u0628\u062a\u062f\u0631\u064a\u0628 \u0646\u0645\u0648\u0630\u062c\u064a\u0646:<\/p>\n<ul class=\"wp-block-list\">\n<li class=\"wp-block-list-item\">\u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0627\u0644\u0623\u0648\u0644\u060c <strong>DeepSeek-R1-Zero<\/strong>\u060c \u0647\u0648 \u0646\u0645\u0648\u0630\u062c \u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u062a\u0645 \u062a\u062f\u0631\u064a\u0628\u0647 \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0627\u0644\u062a\u0639\u0644\u0645 \u0627\u0644\u0645\u0639\u0632\u0632\u060c \u0648\u064a\u0642\u0648\u0645 \u0628\u0625\u0646\u0634\u0627\u0621 \u0628\u064a\u0627\u0646\u0627\u062a \u0644\u062a\u062f\u0631\u064a\u0628 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0627\u0644\u062b\u0627\u0646\u064a\u060c <strong>DeepSeek-R1<\/strong>.<\/li>\n<li class=\"wp-block-list-item\">\u064a\u062d\u0642\u0642 \u0630\u0644\u0643 \u0645\u0646 \u062e\u0644\u0627\u0644 \u0625\u0646\u062a\u0627\u062c \u0622\u062b\u0627\u0631 \u0627\u0633\u062a\u062f\u0644\u0627\u0644\u060c \u064a\u062a\u0645 \u0627\u0644\u0627\u062d\u062a\u0641\u0627\u0638 \u0641\u0642\u0637 \u0628\u0627\u0644\u0645\u062e\u0631\u062c\u0627\u062a \u0639\u0627\u0644\u064a\u0629 \u0627\u0644\u062c\u0648\u062f\u0629 \u0645\u0646\u0647\u0627 \u0628\u0646\u0627\u0621\u064b \u0639\u0644\u0649 \u0646\u062a\u0627\u0626\u062c\u0647\u0627 \u0627\u0644\u0646\u0647\u0627\u0626\u064a\u0629.<\/li>\n<li class=\"wp-block-list-item\">\u0647\u0630\u0627 \u064a\u0639\u0646\u064a \u0623\u0646\u0647 \u0639\u0644\u0649 \u0639\u0643\u0633 \u0645\u0639\u0638\u0645 \u0627\u0644\u0646\u0645\u0627\u0630\u062c\u060c \u0641\u0625\u0646 \u0623\u0645\u062b\u0644\u0629 \u0627\u0644\u062a\u0639\u0644\u0645 \u0627\u0644\u0645\u0639\u0632\u0632 (RL) \u0641\u064a \u0645\u0633\u0627\u0631 \u0627\u0644\u062a\u062f\u0631\u064a\u0628 \u0647\u0630\u0627 \u0644\u0627 \u064a\u062a\u0645 \u062a\u0646\u0633\u064a\u0642\u0647\u0627 \u0628\u0648\u0627\u0633\u0637\u0629 \u0627\u0644\u0628\u0634\u0631 \u0648\u0644\u0643\u0646 \u064a\u062a\u0645 \u0625\u0646\u0634\u0627\u0624\u0647\u0627 \u0628\u0648\u0627\u0633\u0637\u0629 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0646\u0641\u0633\u0647.<\/li>\n<\/ul>\n<p class=\"wp-block-paragraph\">\u0627\u0644\u0646\u062a\u064a\u062c\u0629 \u0647\u064a \u0623\u0646 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u062d\u0642\u0642 \u0623\u062f\u0627\u0621\u064b \u0645\u0645\u0627\u062b\u0644\u0627\u064b \u0644\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0631\u0627\u0626\u062f\u0629 \u0645\u062b\u0644 <a href=\"https:\/\/openai.com\/index\/learning-to-reason-with-llms\/\" target=\"_blank\" rel=\"noreferrer noopener\">\u0646\u0645\u0648\u0630\u062c o1 \u0645\u0646 OpenAI<\/a> \u0641\u064a \u0645\u0647\u0627\u0645 \u0645\u062b\u0644 \u0627\u0644\u0631\u064a\u0627\u0636\u064a\u0627\u062a \u0648\u0627\u0644\u0628\u0631\u0645\u062c\u0629 \u0648\u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0645\u0639\u0642\u062f.<\/p>\n<h2 id=\"heading-three\" class=\"wp-block-heading\">(3) \u0641\u0647\u0645 \u0639\u0645\u0644\u064a\u0629 \u0627\u0644\u062a\u0642\u0637\u064a\u0631 \u0648\u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0645\u064f\u0642\u0637\u0631\u0629 \u0645\u0646 DeepSeek-R1<\/h2>\n<p class=\"wp-block-paragraph\">\u0628\u0627\u0644\u0625\u0636\u0627\u0641\u0629 \u0625\u0644\u0649 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0627\u0644\u0643\u0627\u0645\u0644\u060c \u0642\u0627\u0645\u0648\u0627 \u0623\u064a\u0636\u064b\u0627 \u0628\u0641\u062a\u062d \u0645\u0635\u062f\u0631 \u0633\u062a\u0629 \u0646\u0645\u0627\u0630\u062c \u0643\u062b\u064a\u0641\u0629 \u0623\u0635\u063a\u0631 \u062d\u062c\u0645\u064b\u0627 (\u062a\u0633\u0645\u0649 \u0623\u064a\u0636\u064b\u0627 DeepSeek-R1) \u0628\u0623\u062d\u062c\u0627\u0645 \u0645\u062e\u062a\u0644\u0641\u0629 (1.5B\u060c 7B\u060c 8B\u060c 14B\u060c 32B\u060c 70B)\u060c \u062a\u0645 \u062a\u0642\u0637\u064a\u0631\u0647\u0627 \u0645\u0646 DeepSeek-R1 \u0627\u0633\u062a\u0646\u0627\u062f\u064b\u0627 \u0625\u0644\u0649 <a href=\"https:\/\/huggingface.co\/Qwen\" target=\"_blank\" rel=\"noreferrer noopener\">Qwen<\/a> \u0623\u0648 <a href=\"https:\/\/www.llama.com\/\" target=\"_blank\" rel=\"noreferrer noopener\">Llama<\/a> \u0643\u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0627\u0644\u0623\u0633\u0627\u0633\u064a.<\/p>\n<p class=\"wp-block-paragraph\"><strong>\u0627\u0644\u062a\u0642\u0637\u064a\u0631 (Distillation)<\/strong> \u0647\u0648 \u062a\u0642\u0646\u064a\u0629 \u064a\u062a\u0645 \u0641\u064a\u0647\u0627 \u062a\u062f\u0631\u064a\u0628 \u0646\u0645\u0648\u0630\u062c \u0623\u0635\u063a\u0631 (&#8220;\u0627\u0644\u0637\u0627\u0644\u0628&#8221;) \u0644\u062a\u0643\u0631\u0627\u0631 \u0623\u062f\u0627\u0621 \u0646\u0645\u0648\u0630\u062c \u0623\u0643\u0628\u0631 \u0648\u0623\u0643\u062b\u0631 \u0642\u0648\u0629 \u062a\u0645 \u062a\u062f\u0631\u064a\u0628\u0647 \u0645\u0633\u0628\u0642\u064b\u0627 (&#8220;\u0627\u0644\u0645\u0639\u0644\u0645&#8221;).<\/p>\n<p><img decoding=\"async\" class=\"wp-image-601944 aligncenter\" src=\"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2025\/05\/Untitled-1024x510-1.png\" alt=\"\" \/ width=\"1024\" height=\"510\"><\/p>\n<p class=\"wp-block-paragraph\">\u0641\u064a \u0647\u0630\u0647 \u0627\u0644\u062d\u0627\u0644\u0629\u060c \u0627\u0644\u0645\u064f\u0639\u0644\u0645 \u0647\u0648 \u0646\u0645\u0648\u0630\u062c DeepSeek-R1 \u0628\u062d\u062c\u0645 671B\u060c \u0648\u0627\u0644\u0637\u0644\u0627\u0628 \u0647\u0645 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0633\u062a\u0629 \u0627\u0644\u062a\u064a \u062a\u0645 \u062a\u0642\u0637\u064a\u0631\u0647\u0627 \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0647\u0630\u0627 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0627\u0644\u0623\u0633\u0627\u0633\u064a \u0645\u0641\u062a\u0648\u062d \u0627\u0644\u0645\u0635\u062f\u0631:<\/p>\n<ul class=\"wp-block-list\">\n<li class=\"wp-block-list-item\"><a href=\"https:\/\/huggingface.co\/deepseek-ai\/DeepSeek-R1-Distill-Qwen-1.5B\" target=\"_blank\" rel=\"noreferrer noopener\">Qwen2.5\u200a\u2014\u200aMath-1.5B<\/a><\/li>\n<li class=\"wp-block-list-item\"><a href=\"https:\/\/huggingface.co\/deepseek-ai\/DeepSeek-R1-Distill-Qwen-7B\" target=\"_blank\" rel=\"noreferrer noopener\">Qwen2.5\u200a\u2014\u200aMath-7B<\/a><\/li>\n<li class=\"wp-block-list-item\"><a href=\"https:\/\/huggingface.co\/deepseek-ai\/DeepSeek-R1-Distill-Qwen-14B\" target=\"_blank\" rel=\"noreferrer noopener\">Qwen2.5\u200a\u2014\u200a14B<\/a><\/li>\n<li class=\"wp-block-list-item\"><a href=\"https:\/\/huggingface.co\/deepseek-ai\/DeepSeek-R1-Distill-Qwen-32B\" target=\"_blank\" rel=\"noreferrer noopener\">Qwen2.5\u200a\u2014\u200a32B<\/a><\/li>\n<li class=\"wp-block-list-item\"><a href=\"https:\/\/huggingface.co\/deepseek-ai\/DeepSeek-R1-Distill-Llama-8B\" target=\"_blank\" rel=\"noreferrer noopener\">Llama-3.1\u200a\u2014\u200a8B<\/a><\/li>\n<li class=\"wp-block-list-item\"><a href=\"https:\/\/huggingface.co\/deepseek-ai\/DeepSeek-R1-Distill-Llama-70B\" target=\"_blank\" rel=\"noreferrer noopener\">Llama-3.3\u200a\u2014\u200a70B-Instruct<\/a><\/li>\n<\/ul>\n<p class=\"wp-block-paragraph\">\u062a\u0645 \u0627\u0633\u062a\u062e\u062f\u0627\u0645 DeepSeek-R1 \u0643\u0646\u0645\u0648\u0630\u062c \u0645\u0639\u0644\u0645 \u0644\u0625\u0646\u0634\u0627\u0621 800,000 \u0639\u064a\u0646\u0629 \u062a\u062f\u0631\u064a\u0628\u060c \u0648\u0647\u064a \u0645\u0632\u064a\u062c \u0645\u0646 \u0639\u064a\u0646\u0627\u062a \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0648\u063a\u064a\u0631 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644\u060c \u0644\u0644\u062a\u0642\u0637\u064a\u0631 \u0639\u0628\u0631 <strong>\u0627\u0644\u0636\u0628\u0637 \u0627\u0644\u062f\u0642\u064a\u0642 \u0627\u0644\u062e\u0627\u0636\u0639 \u0644\u0644\u0625\u0634\u0631\u0627\u0641 (supervised fine-tuning)<\/strong> \u0644\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0623\u0633\u0627\u0633\u064a\u0629 (1.5B\u060c 7B\u060c 8B\u060c 14B\u060c 32B\u060c \u0648 70B).<\/p>\n<p class=\"wp-block-paragraph\"><strong>\u0625\u0630\u064b\u0627\u060c \u0644\u0645\u0627\u0630\u0627 \u0646\u0642\u0648\u0645 \u0628\u0627\u0644\u062a\u0642\u0637\u064a\u0631 \u0641\u064a \u0627\u0644\u0645\u0642\u0627\u0645 \u0627\u0644\u0623\u0648\u0644\u061f<\/strong><\/p>\n<p class=\"wp-block-paragraph\">\u0627\u0644\u0647\u062f\u0641 \u0647\u0648 \u0646\u0642\u0644 \u0642\u062f\u0631\u0627\u062a \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0644\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0623\u0643\u0628\u0631\u060c \u0645\u062b\u0644 DeepSeek-R1 671B\u060c \u0625\u0644\u0649 \u0646\u0645\u0627\u0630\u062c \u0623\u0635\u063a\u0631 \u0648\u0623\u0643\u062b\u0631 \u0643\u0641\u0627\u0621\u0629. \u0648\u0647\u0630\u0627 \u064a\u0645\u0643\u0651\u0646 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0623\u0635\u063a\u0631 \u0645\u0646 \u0627\u0644\u062a\u0639\u0627\u0645\u0644 \u0645\u0639 \u0645\u0647\u0627\u0645 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0645\u0639\u0642\u062f\u0629 \u0645\u0639 \u0643\u0648\u0646\u0647\u0627 \u0623\u0633\u0631\u0639 \u0648\u0623\u0643\u062b\u0631 \u0643\u0641\u0627\u0621\u0629 \u0641\u064a \u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0627\u0644\u0645\u0648\u0627\u0631\u062f.<\/p>\n<p class=\"wp-block-paragraph\">\u0639\u0644\u0627\u0648\u0629 \u0639\u0644\u0649 \u0630\u0644\u0643\u060c \u064a\u062d\u062a\u0648\u064a DeepSeek-R1 \u0639\u0644\u0649 \u0639\u062f\u062f \u0647\u0627\u0626\u0644 \u0645\u0646 \u0627\u0644\u0645\u0639\u0644\u0645\u0627\u062a (671 \u0645\u0644\u064a\u0627\u0631)\u060c \u0645\u0645\u0627 \u064a\u062c\u0639\u0644 \u062a\u0634\u063a\u064a\u0644\u0647 \u0623\u0645\u0631\u064b\u0627 \u0635\u0639\u0628\u064b\u0627 \u0639\u0644\u0649 \u0645\u0639\u0638\u0645 \u0627\u0644\u0623\u062c\u0647\u0632\u0629 \u0627\u0644\u0627\u0633\u062a\u0647\u0644\u0627\u0643\u064a\u0629.<\/p>\n<p class=\"wp-block-paragraph\">\u062d\u062a\u0649 \u0623\u0642\u0648\u0649 \u062c\u0647\u0627\u0632 MacBook Pro\u060c \u0628\u0630\u0627\u0643\u0631\u0629 \u0645\u0648\u062d\u062f\u0629 \u0642\u0635\u0648\u0649 \u062a\u0628\u0644\u063a 128GB\u060c \u063a\u064a\u0631 \u0643\u0627\u0641\u064d \u0644\u062a\u0634\u063a\u064a\u0644 \u0646\u0645\u0648\u0630\u062c \u0628\u0645\u0639\u0644\u0645\u0627\u062a \u062a\u0628\u0644\u063a 671 \u0645\u0644\u064a\u0627\u0631.<\/p>\n<p class=\"wp-block-paragraph\">\u0639\u0644\u0649 \u0647\u0630\u0627 \u0627\u0644\u0646\u062d\u0648\u060c \u062a\u0641\u062a\u062d \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0625\u0645\u0643\u0627\u0646\u064a\u0629 \u0646\u0634\u0631\u0647\u0627 \u0639\u0644\u0649 \u0627\u0644\u0623\u062c\u0647\u0632\u0629 \u0630\u0627\u062a \u0627\u0644\u0645\u0648\u0627\u0631\u062f \u0627\u0644\u062d\u0633\u0627\u0628\u064a\u0629 \u0627\u0644\u0645\u062d\u062f\u0648\u062f\u0629.<\/p>\n<blockquote><p>\u062d\u0642\u0642\u062a <a href=\"https:\/\/unsloth.ai\/blog\/deepseekr1-dynamic\" target=\"_blank\" rel=\"noreferrer noopener\">Unsloth<\/a> \u0625\u0646\u062c\u0627\u0632\u064b\u0627 \u0628\u0627\u0647\u0631\u064b\u0627 \u0639\u0646 \u0637\u0631\u064a\u0642 \u062a\u0643\u0645\u064a\u0645 \u0646\u0645\u0648\u0630\u062c DeepSeek-R1 \u0627\u0644\u0623\u0635\u0644\u064a \u0630\u064a \u0627\u0644\u0640 671 \u0645\u0644\u064a\u0627\u0631 \u0645\u0639\u0644\u0645\u0629 \u0625\u0644\u0649 131 \u062c\u064a\u062c\u0627\u0628\u0627\u064a\u062a \u0641\u0642\u0637 &#8211; \u0648\u0647\u0648 \u062a\u062e\u0641\u064a\u0636 \u0645\u0644\u062d\u0648\u0638 \u0628\u0646\u0633\u0628\u0629 80% \u0641\u064a \u0627\u0644\u062d\u062c\u0645. \u0648\u0645\u0639 \u0630\u0644\u0643\u060c \u064a\u0638\u0644 \u0645\u062a\u0637\u0644\u0628 \u0630\u0627\u0643\u0631\u0629 \u0627\u0644\u0648\u0635\u0648\u0644 \u0627\u0644\u0639\u0634\u0648\u0627\u0626\u064a (VRAM) \u0628\u0633\u0639\u0629 131 \u062c\u064a\u062c\u0627\u0628\u0627\u064a\u062a \u064a\u0634\u0643\u0644 \u0639\u0627\u0626\u0642\u064b\u0627 \u0643\u0628\u064a\u0631\u064b\u0627\u060c \u062e\u0627\u0635\u0629 \u0644\u0644\u0645\u0637\u0648\u0631\u064a\u0646 \u0627\u0644\u0630\u064a\u0646 \u064a\u0639\u0645\u0644\u0648\u0646 \u0639\u0644\u0649 \u0623\u062c\u0647\u0632\u0629 \u0630\u0627\u062a \u0645\u0648\u0627\u0631\u062f \u0645\u062d\u062f\u0648\u062f\u0629. \u0647\u0630\u0627 \u0627\u0644\u0625\u0646\u062c\u0627\u0632 \u064a\u0645\u062b\u0644 \u062e\u0637\u0648\u0629 \u0647\u0627\u0645\u0629 \u0646\u062d\u0648 \u062c\u0639\u0644 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0627\u0644\u0643\u0628\u064a\u0631\u0629 \u0641\u064a \u0645\u062a\u0646\u0627\u0648\u0644 \u0634\u0631\u064a\u062d\u0629 \u0623\u0648\u0633\u0639 \u0645\u0646 \u0627\u0644\u0645\u0633\u062a\u062e\u062f\u0645\u064a\u0646.<\/p><\/blockquote>\n<h2 id=\"heading-four\" class=\"wp-block-heading\">(4) \u0627\u062e\u062a\u064a\u0627\u0631 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0627\u0644\u0645\u064f\u0642\u0637\u0651\u064e\u0631 \u0627\u0644\u0623\u0645\u062b\u0644<\/h2>\n<p class=\"wp-block-paragraph\">\u0645\u0639 \u062a\u0648\u0641\u0631 \u0633\u062a\u0629 \u0623\u062d\u062c\u0627\u0645 \u0645\u062e\u062a\u0644\u0641\u0629 \u0644\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0645\u064f\u0642\u0637\u0651\u064e\u0631\u0629 \u0644\u0644\u0627\u062e\u062a\u064a\u0627\u0631 \u0645\u0646 \u0628\u064a\u0646\u0647\u0627\u060c \u064a\u0639\u062a\u0645\u062f \u062a\u062d\u062f\u064a\u062f \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0627\u0644\u0645\u0646\u0627\u0633\u0628 \u0625\u0644\u0649 \u062d\u062f \u0643\u0628\u064a\u0631 \u0639\u0644\u0649 \u0642\u062f\u0631\u0627\u062a \u0627\u0644\u0623\u062c\u0647\u0632\u0629 \u0627\u0644\u0645\u062d\u0644\u064a\u0629.<\/p>\n<p class=\"wp-block-paragraph\">\u0628\u0627\u0644\u0646\u0633\u0628\u0629 \u0644\u0623\u0648\u0644\u0626\u0643 \u0627\u0644\u0630\u064a\u0646 \u064a\u0645\u062a\u0644\u0643\u0648\u0646 \u0648\u062d\u062f\u0627\u062a \u0645\u0639\u0627\u0644\u062c\u0629 \u0631\u0633\u0648\u0645\u064a\u0629 (GPUs) \u0623\u0648 \u0648\u062d\u062f\u0627\u062a \u0645\u0639\u0627\u0644\u062c\u0629 \u0645\u0631\u0643\u0632\u064a\u0629 (CPUs) \u0639\u0627\u0644\u064a\u0629 \u0627\u0644\u0623\u062f\u0627\u0621 \u0648\u064a\u062d\u062a\u0627\u062c\u0648\u0646 \u0625\u0644\u0649 \u0623\u0642\u0635\u0649 \u0642\u062f\u0631 \u0645\u0646 \u0627\u0644\u0623\u062f\u0627\u0621\u060c \u062a\u064f\u0639\u062f \u0646\u0645\u0627\u0630\u062c DeepSeek-R1 \u0627\u0644\u0623\u0643\u0628\u0631 (32B \u0648\u0645\u0627 \u0641\u0648\u0642) \u0645\u062b\u0627\u0644\u064a\u0629 &#8211; \u062d\u062a\u0649 \u0627\u0644\u0625\u0635\u062f\u0627\u0631 \u0627\u0644\u0643\u0645\u064a 671B \u064a\u0639\u062a\u0628\u0631 \u0642\u0627\u0628\u0644\u0627\u064b \u0644\u0644\u062a\u0637\u0628\u064a\u0642.<\/p>\n<p class=\"wp-block-paragraph\">\u0648\u0645\u0639 \u0630\u0644\u0643\u060c \u0625\u0630\u0627 \u0643\u0627\u0646\u062a \u0627\u0644\u0645\u0648\u0627\u0631\u062f \u0645\u062d\u062f\u0648\u062f\u0629 \u0623\u0648 \u0643\u0646\u062a \u062a\u0641\u0636\u0644 \u0623\u0648\u0642\u0627\u062a \u0625\u0646\u0634\u0627\u0621 \u0623\u0633\u0631\u0639 (\u0643\u0645\u0627 \u0647\u0648 \u0627\u0644\u062d\u0627\u0644 \u0628\u0627\u0644\u0646\u0633\u0628\u0629 \u0644\u064a)\u060c \u0641\u0625\u0646 \u0627\u0644\u0645\u062a\u063a\u064a\u0631\u0627\u062a \u0627\u0644\u0645\u064f\u0642\u0637\u0651\u064e\u0631\u0629 \u0627\u0644\u0623\u0635\u063a\u0631\u060c \u0645\u062b\u0644 8B \u0623\u0648 14B\u060c \u062a\u0639\u062a\u0628\u0631 \u062e\u064a\u0627\u0631\u064b\u0627 \u0623\u0641\u0636\u0644. \u0647\u0630\u0627 \u064a\u0648\u0627\u0632\u0646 \u0628\u064a\u0646 \u0627\u0644\u0623\u062f\u0627\u0621 \u0648\u0645\u062a\u0637\u0644\u0628\u0627\u062a \u0627\u0644\u0645\u0648\u0627\u0631\u062f.<\/p>\n<p class=\"wp-block-paragraph\"><strong>\u0644\u0647\u0630\u0627 \u0627\u0644\u0645\u0634\u0631\u0648\u0639\u060c \u0633\u0623\u0633\u062a\u062e\u062f\u0645 \u0646\u0645\u0648\u0630\u062c DeepSeek-R1 \u0627\u0644\u0645\u064f\u0642\u0637\u0651\u064e\u0631 <\/strong><a href=\"https:\/\/huggingface.co\/deepseek-ai\/DeepSeek-R1-Distill-Qwen-14B\" target=\"_blank\" rel=\"noreferrer noopener\"><strong>Qwen-14B<\/strong><\/a><strong>\u060c \u0648\u0627\u0644\u0630\u064a \u064a\u062a\u0648\u0627\u0641\u0642 \u0645\u0639 \u0642\u064a\u0648\u062f \u0627\u0644\u0623\u062c\u0647\u0632\u0629 \u0627\u0644\u062a\u064a \u0648\u0627\u062c\u0647\u062a\u0647\u0627.<\/strong> \u0647\u0630\u0627 \u0627\u0644\u0646\u0645\u0648\u0630\u062c (14B) \u064a\u0645\u062b\u0644 \u062d\u0644\u0627\u064b \u0648\u0633\u0637\u064b\u0627 \u0645\u0645\u062a\u0627\u0632\u064b\u0627 \u0628\u064a\u0646 \u0627\u0644\u062f\u0642\u0629 \u0648\u0627\u0644\u0633\u0631\u0639\u0629\u060c \u0645\u0645\u0627 \u064a\u062c\u0639\u0644\u0647 \u0645\u0646\u0627\u0633\u0628\u064b\u0627 \u062a\u0645\u0627\u0645\u064b\u0627 \u0644\u0628\u064a\u0626\u0629 \u0627\u0644\u062a\u0637\u0648\u064a\u0631 \u0627\u0644\u062e\u0627\u0635\u0629 \u0628\u064a.<\/p>\n<h2 id=\"heading-five\" class=\"wp-block-heading\">(5) \u0645\u0639\u0627\u064a\u064a\u0631 \u062a\u0642\u064a\u064a\u0645 \u0627\u0644\u0642\u062f\u0631\u0629 \u0639\u0644\u0649 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0644\u062f\u0649 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0627\u062a \u0627\u0644\u0643\u0628\u064a\u0631\u0629<\/h2>\n<p class=\"wp-block-paragraph\">\u0639\u0627\u062f\u0629\u064b \u0645\u0627 \u064a\u062a\u0645 \u062a\u0642\u064a\u064a\u0645 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0627\u062a \u0627\u0644\u0643\u0628\u064a\u0631\u0629 (LLMs) \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0645\u0639\u0627\u064a\u064a\u0631 \u0642\u064a\u0627\u0633\u064a\u0629 \u062a\u062d\u062f\u062f \u0645\u0633\u062a\u0648\u0649 \u0623\u062f\u0627\u0626\u0647\u0627 \u0641\u064a \u0645\u0647\u0627\u0645 \u0645\u062a\u0646\u0648\u0639\u0629\u060c \u0628\u0645\u0627 \u0641\u064a \u0630\u0644\u0643 \u0641\u0647\u0645 \u0627\u0644\u0644\u063a\u0629\u060c \u0648\u062a\u0648\u0644\u064a\u062f \u0627\u0644\u062a\u0639\u0644\u064a\u0645\u0627\u062a \u0627\u0644\u0628\u0631\u0645\u062c\u064a\u0629\u060c \u0648\u0627\u062a\u0628\u0627\u0639 \u0627\u0644\u062a\u0639\u0644\u064a\u0645\u0627\u062a\u060c \u0648\u0627\u0644\u0625\u062c\u0627\u0628\u0629 \u0639\u0644\u0649 \u0627\u0644\u0623\u0633\u0626\u0644\u0629. \u062a\u062a\u0636\u0645\u0646 \u0627\u0644\u0623\u0645\u062b\u0644\u0629 \u0627\u0644\u0634\u0627\u0626\u0639\u0629 \u0645\u0642\u0627\u064a\u064a\u0633 \u0645\u062b\u0644 <a href=\"https:\/\/paperswithcode.com\/sota\/multi-task-language-understanding-on-mmlu\" target=\"_blank\" rel=\"noreferrer noopener\">MMLU<\/a>\u060c \u0648 <a href=\"https:\/\/paperswithcode.com\/sota\/code-generation-on-humaneval\" target=\"_blank\" rel=\"noreferrer noopener\">HumanEval<\/a>\u060c \u0648 <a href=\"https:\/\/paperswithcode.com\/dataset\/mgsm\" target=\"_blank\" rel=\"noreferrer noopener\">MGSM<\/a>. \u062a\u0639\u062a\u0628\u0631 \u0647\u0630\u0647 \u0627\u0644\u0645\u0642\u0627\u064a\u064a\u0633 \u0623\u0633\u0627\u0633\u064a\u0629 \u0641\u064a \u062a\u0642\u064a\u064a\u0645 \u0642\u062f\u0631\u0627\u062a \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0627\u062a \u0627\u0644\u0643\u0628\u064a\u0631\u0629.<\/p>\n<p class=\"wp-block-paragraph\">\u0644\u0642\u064a\u0627\u0633 \u0642\u062f\u0631\u0629 \u0646\u0645\u0648\u0630\u062c \u0644\u063a\u0648\u064a \u0643\u0628\u064a\u0631 \u0639\u0644\u0649 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0645\u0646\u0637\u0642\u064a\u060c \u0646\u062d\u062a\u0627\u062c \u0625\u0644\u0649 \u0645\u0639\u0627\u064a\u064a\u0631 \u0623\u0643\u062b\u0631 \u062a\u062d\u062f\u064a\u064b\u0627 \u0648\u062a\u0631\u0643\u0632 \u0628\u0634\u0643\u0644 \u0643\u0628\u064a\u0631 \u0639\u0644\u0649 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644\u060c \u0648\u062a\u062a\u062c\u0627\u0648\u0632 \u0627\u0644\u0645\u0647\u0627\u0645 \u0627\u0644\u0633\u0637\u062d\u064a\u0629. \u0641\u064a\u0645\u0627 \u064a\u0644\u064a \u0628\u0639\u0636 \u0627\u0644\u0623\u0645\u062b\u0644\u0629 \u0627\u0644\u0634\u0627\u0626\u0639\u0629 \u0627\u0644\u062a\u064a \u062a\u0631\u0643\u0632 \u0639\u0644\u0649 \u062a\u0642\u064a\u064a\u0645 \u0642\u062f\u0631\u0627\u062a \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0645\u062a\u0642\u062f\u0645\u0629:<\/p>\n<h3 class=\"wp-block-heading\">(i) \u0627\u0645\u062a\u062d\u0627\u0646 AIME 2024: \u0627\u0644\u0631\u064a\u0627\u0636\u064a\u0627\u062a \u0627\u0644\u062a\u0646\u0627\u0641\u0633\u064a\u0629<\/h3>\n<ul class=\"wp-block-list\">\n<li class=\"wp-block-list-item\">\u064a\u064f\u0639\u062f <a href=\"https:\/\/paperswithcode.com\/sota\/mathematical-reasoning-on-aime24\" target=\"_blank\" rel=\"noreferrer noopener\"><strong>\u0627\u0645\u062a\u062d\u0627\u0646 \u0627\u0644\u062f\u0639\u0648\u0629 \u0627\u0644\u0623\u0645\u0631\u064a\u0643\u064a \u0644\u0644\u0631\u064a\u0627\u0636\u064a\u0627\u062a (AIME) 2024<\/strong><\/a> \u0645\u0639\u064a\u0627\u0631\u064b\u0627 \u0642\u0648\u064a\u064b\u0627 \u0644\u062a\u0642\u064a\u064a\u0645 \u0642\u062f\u0631\u0627\u062a \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0648\u064a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631\u0629 (LLM) \u0641\u064a \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0631\u064a\u0627\u0636\u064a.<\/li>\n<li class=\"wp-block-list-item\">\u064a\u0645\u062b\u0644 \u0647\u0630\u0627 \u0627\u0644\u0627\u0645\u062a\u062d\u0627\u0646 \u062a\u062d\u062f\u064a\u064b\u0627 \u0643\u0628\u064a\u0631\u064b\u0627 \u0641\u064a \u0645\u062c\u0627\u0644 \u0627\u0644\u0631\u064a\u0627\u0636\u064a\u0627\u062a \u0627\u0644\u062a\u0646\u0627\u0641\u0633\u064a\u0629\u060c \u062d\u064a\u062b \u064a\u0637\u0631\u062d \u0645\u0633\u0627\u0626\u0644 \u0645\u0639\u0642\u062f\u0629 \u0648\u0645\u062a\u0639\u062f\u062f\u0629 \u0627\u0644\u062e\u0637\u0648\u0627\u062a. \u064a\u062e\u062a\u0628\u0631 \u0627\u0644\u0627\u0645\u062a\u062d\u0627\u0646 \u0642\u062f\u0631\u0629 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0648\u064a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631\u0629 \u0639\u0644\u0649 \u0641\u0647\u0645 \u0627\u0644\u0623\u0633\u0626\u0644\u0629 \u0627\u0644\u0645\u0639\u0642\u062f\u0629\u060c \u0648\u062a\u0637\u0628\u064a\u0642 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0645\u062a\u0642\u062f\u0645\u060c \u0648\u0625\u062c\u0631\u0627\u0621 \u0639\u0645\u0644\u064a\u0627\u062a \u0627\u0644\u062a\u0644\u0627\u0639\u0628 \u0627\u0644\u0631\u0645\u0632\u064a \u0627\u0644\u062f\u0642\u064a\u0642\u0629. \u064a\u0639\u062a\u0628\u0631 AIME \u0645\u0642\u064a\u0627\u0633\u064b\u0627 \u0647\u0627\u0645\u064b\u0627 \u0644\u062a\u0642\u064a\u064a\u0645 \u0645\u0647\u0627\u0631\u0627\u062a \u062d\u0644 \u0627\u0644\u0645\u0634\u0643\u0644\u0627\u062a \u0627\u0644\u0631\u064a\u0627\u0636\u064a\u0629 \u0627\u0644\u0645\u0639\u0642\u062f\u0629.<\/li>\n<\/ul>\n<h3 class=\"wp-block-heading\">(ii) Codeforces &#8211; \u0643\u0648\u062f \u0627\u0644\u0645\u0646\u0627\u0641\u0633\u0629<\/h3>\n<ul class=\"wp-block-list\">\n<li class=\"wp-block-list-item\">\u064a\u0642\u0648\u0645 <a href=\"https:\/\/arxiv.org\/abs\/2501.01257v2\" target=\"_blank\" rel=\"noreferrer noopener\"><strong>\u0645\u0639\u064a\u0627\u0631 Codeforces<\/strong><\/a> \u0628\u062a\u0642\u064a\u064a\u0645 \u0642\u062f\u0631\u0629 \u0646\u0645\u0648\u0630\u062c \u0627\u0644\u0644\u063a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631 (LLM) \u0639\u0644\u0649 \u0627\u0644\u0627\u0633\u062a\u0646\u062a\u0627\u062c \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0645\u0633\u0627\u0626\u0644 \u0628\u0631\u0645\u062c\u0629 \u062a\u0646\u0627\u0641\u0633\u064a\u0629 \u062d\u0642\u064a\u0642\u064a\u0629 \u0645\u0646 Codeforces\u060c \u0648\u0647\u064a \u0645\u0646\u0635\u0629 \u0645\u0639\u0631\u0648\u0641\u0629 \u0628\u062a\u062d\u062f\u064a\u0627\u062a \u0627\u0644\u062e\u0648\u0627\u0631\u0632\u0645\u064a\u0627\u062a. \u064a\u0639\u062a\u0628\u0631 Codeforces \u0645\u0639\u064a\u0627\u0631\u064b\u0627 \u0630\u0647\u0628\u064a\u064b\u0627 \u0644\u062a\u0642\u064a\u064a\u0645 \u0642\u062f\u0631\u0627\u062a \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0641\u064a \u062d\u0644 \u0627\u0644\u0645\u0634\u0643\u0644\u0627\u062a \u0627\u0644\u0645\u0639\u0642\u062f\u0629.<\/li>\n<li class=\"wp-block-list-item\">\u062a\u062e\u062a\u0628\u0631 \u0647\u0630\u0647 \u0627\u0644\u0645\u0634\u0643\u0644\u0627\u062a \u0642\u062f\u0631\u0629 \u0646\u0645\u0648\u0630\u062c \u0627\u0644\u0644\u063a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631 (LLM) \u0639\u0644\u0649 \u0641\u0647\u0645 \u0627\u0644\u062a\u0639\u0644\u064a\u0645\u0627\u062a \u0627\u0644\u0645\u0639\u0642\u062f\u0629\u060c \u0648\u0625\u062c\u0631\u0627\u0621 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0645\u0646\u0637\u0642\u064a \u0648\u0627\u0644\u0631\u064a\u0627\u0636\u064a\u060c \u0648\u062a\u062e\u0637\u064a\u0637 \u062d\u0644\u0648\u0644 \u0645\u062a\u0639\u062f\u062f\u0629 \u0627\u0644\u062e\u0637\u0648\u0627\u062a\u060c \u0648\u0625\u0646\u0634\u0627\u0621 \u062a\u0639\u0644\u064a\u0645\u0627\u062a \u0628\u0631\u0645\u062c\u064a\u0629 \u0635\u062d\u064a\u062d\u0629 \u0648\u0641\u0639\u0627\u0644\u0629. \u062a\u062a\u0637\u0644\u0628 \u0647\u0630\u0647 \u0627\u0644\u0645\u0633\u0627\u0626\u0644 \u0641\u0647\u0645\u064b\u0627 \u0639\u0645\u064a\u0642\u064b\u0627 \u0644\u0644\u062e\u0648\u0627\u0631\u0632\u0645\u064a\u0627\u062a \u0648\u0647\u064a\u0627\u0643\u0644 \u0627\u0644\u0628\u064a\u0627\u0646\u0627\u062a\u060c \u0628\u0627\u0644\u0625\u0636\u0627\u0641\u0629 \u0625\u0644\u0649 \u0627\u0644\u0642\u062f\u0631\u0629 \u0639\u0644\u0649 \u062a\u0631\u062c\u0645\u0629 \u0627\u0644\u0645\u0634\u0643\u0644\u0629 \u0625\u0644\u0649 \u0643\u0648\u062f \u0642\u0627\u0628\u0644 \u0644\u0644\u062a\u0646\u0641\u064a\u0630.<\/li>\n<\/ul>\n<h3 class=\"wp-block-heading\">(iii) GPQA Diamond &#8211; \u0623\u0633\u0626\u0644\u0629 \u0639\u0644\u0645\u064a\u0629 \u0628\u0645\u0633\u062a\u0648\u0649 \u0627\u0644\u062f\u0643\u062a\u0648\u0631\u0627\u0647<\/h3>\n<ul class=\"wp-block-list\">\n<li class=\"wp-block-list-item\">GPQA-Diamond \u0647\u064a \u0645\u062c\u0645\u0648\u0639\u0629 \u0641\u0631\u0639\u064a\u0629 \u0645\u064f\u0646\u062a\u0642\u0627\u0629 \u0645\u0646 <strong>\u0623\u0635\u0639\u0628 \u0627\u0644\u0623\u0633\u0626\u0644\u0629<\/strong> \u0645\u0646 \u0645\u0639\u064a\u0627\u0631 <a href=\"https:\/\/arxiv.org\/abs\/2311.12022\" target=\"_blank\" rel=\"noreferrer noopener\"><strong>GPQA (\u0625\u062c\u0627\u0628\u0629 \u0623\u0633\u0626\u0644\u0629 \u0627\u0644\u0641\u064a\u0632\u064a\u0627\u0621 \u0639\u0644\u0649 \u0645\u0633\u062a\u0648\u0649 \u0627\u0644\u062f\u0631\u0627\u0633\u0627\u062a \u0627\u0644\u0639\u0644\u064a\u0627)<\/strong><\/a> \u0627\u0644\u0623\u0648\u0633\u0639\u060c \u0648\u0647\u064a \u0645\u064f\u0635\u0645\u0645\u0629 \u062e\u0635\u064a\u0635\u064b\u0627 \u0644\u062a\u0648\u0633\u064a\u0639 \u062d\u062f\u0648\u062f \u0642\u062f\u0631\u0629 \u0646\u0645\u0627\u0630\u062c LLM \u0639\u0644\u0649 \u0627\u0644\u0627\u0633\u062a\u0646\u062a\u0627\u062c \u0641\u064a \u0645\u0648\u0636\u0648\u0639\u0627\u062a \u0645\u062a\u0642\u062f\u0645\u0629 \u0628\u0645\u0633\u062a\u0648\u0649 \u0627\u0644\u062f\u0643\u062a\u0648\u0631\u0627\u0647. \u064a\u0639\u062a\u0628\u0631 \u0647\u0630\u0627 \u0627\u0644\u0645\u0639\u064a\u0627\u0631 \u062a\u062d\u062f\u064a\u064b\u0627 \u062d\u0642\u064a\u0642\u064a\u064b\u0627 \u0644\u0642\u062f\u0631\u0627\u062a \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0641\u064a \u0641\u0647\u0645 \u0648\u0627\u0633\u062a\u0646\u062a\u0627\u062c \u0627\u0644\u0645\u0641\u0627\u0647\u064a\u0645 \u0627\u0644\u0639\u0644\u0645\u064a\u0629 \u0627\u0644\u0645\u0639\u0642\u062f\u0629.<\/li>\n<li class=\"wp-block-list-item\">\u0628\u064a\u0646\u0645\u0627 \u064a\u0634\u062a\u0645\u0644 GPQA \u0639\u0644\u0649 \u0645\u062c\u0645\u0648\u0639\u0629 \u0645\u0646 \u0623\u0633\u0626\u0644\u0629 \u0627\u0644\u062f\u0631\u0627\u0633\u0627\u062a \u0627\u0644\u0639\u0644\u064a\u0627 \u0627\u0644\u0645\u0641\u0627\u0647\u064a\u0645\u064a\u0629 \u0648\u0627\u0644\u062a\u064a \u062a\u0639\u062a\u0645\u062f \u0639\u0644\u0649 \u0627\u0644\u0639\u0645\u0644\u064a\u0627\u062a \u0627\u0644\u062d\u0633\u0627\u0628\u064a\u0629\u060c \u0641\u0625\u0646 GPQA-Diamond \u064a\u0639\u0632\u0644 \u0641\u0642\u0637 \u0627\u0644\u0623\u0633\u0626\u0644\u0629 \u0627\u0644\u0623\u0643\u062b\u0631 \u062a\u062d\u062f\u064a\u064b\u0627 \u0648\u062a\u0644\u0643 \u0627\u0644\u062a\u064a \u062a\u062a\u0637\u0644\u0628 \u0627\u0633\u062a\u0646\u062a\u0627\u062c\u064b\u0627 \u0645\u0643\u062b\u0641\u064b\u0627.<\/li>\n<li class=\"wp-block-list-item\">\u064a\u0639\u062a\u0628\u0631 \u0647\u0630\u0627 \u0627\u0644\u0645\u0639\u064a\u0627\u0631 &#8220;\u0645\u0642\u0627\u0648\u0645\u064b\u0627 \u0644\u0640 Google&#8221;\u060c \u0645\u0645\u0627 \u064a\u0639\u0646\u064a \u0623\u0646\u0647 \u0645\u0646 \u0627\u0644\u0635\u0639\u0628 \u0627\u0644\u0625\u062c\u0627\u0628\u0629 \u0639\u0644\u064a\u0647 \u062d\u062a\u0649 \u0645\u0639 \u0627\u0644\u0648\u0635\u0648\u0644 \u063a\u064a\u0631 \u0627\u0644\u0645\u0642\u064a\u062f \u0625\u0644\u0649 \u0627\u0644\u0648\u064a\u0628. \u0648\u0647\u0630\u0627 \u064a\u062c\u0639\u0644\u0647 \u0623\u062f\u0627\u0629 \u0642\u064a\u0645\u0629 \u0644\u062a\u0642\u064a\u064a\u0645 \u0642\u062f\u0631\u0629 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0648\u064a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631\u0629 \u0639\u0644\u0649 \u0627\u0644\u062a\u0641\u0643\u064a\u0631 \u0628\u0634\u0643\u0644 \u0645\u0633\u062a\u0642\u0644.<\/li>\n<li class=\"wp-block-list-item\">\u0641\u064a\u0645\u0627 \u064a\u0644\u064a \u0645\u062b\u0627\u0644 \u0639\u0644\u0649 \u0633\u0624\u0627\u0644 GPQA-Diamond:<\/li>\n<\/ul>\n<pre>### GPQA Diamond - Example Question (Molecular Biology)\r\n\r\nA eukaryotic cell evolved a mechanism to turn macromolecular building blocks into energy. \r\nThe process occurs in mitochondria, which are cellular energy factories.\r\n\r\nIn the series of redox reactions, the energy from food is stored between the phosphate groups and used as a universal cellular currency. \r\nThe energy-laden molecules are shuttled out of the mitochondrion to serve in all cellular processes.\r\n\r\nYou discovered a new anti-diabetes drug and want to investigate whether it has an effect on the mitochondria. \r\nYou set up a range of experiments with your HEK293 cell line.\r\n\r\nWhich of the experiments listed below will not help you discover the mitochondrial role of your drug:\r\n\r\n(A) Differential centrifugation extraction of mitochondria followed by the Glucose Uptake Colorimetric Assay Kit \r\n(B) Flow cytometry after labeling with 2.5 \u00b5M 5,5',6,6'-Tetrachloro-1,1',3,3'-tetraethylbenzimidazolylcarbocyanine iodide \r\n(C) Transformation of cells with recombinant luciferase and luminometer reading after 5 \u00b5M of luciferin addition to the supernatant \r\n(D) Confocal fluorescence microscopy after Mito-RTP staining of the cells<\/pre>\n<p class=\"wp-block-paragraph\">\u0641\u064a \u0647\u0630\u0627 \u0627\u0644\u0645\u0634\u0631\u0648\u0639\u060c <strong>\u0646\u0633\u062a\u062e\u062f\u0645 GPQA-Diamond \u0643\u0645\u0639\u064a\u0627\u0631 \u0644\u0644\u0627\u0633\u062a\u0646\u062a\u0627\u062c<\/strong>\u060c \u0643\u0645\u0627 \u0627\u0633\u062a\u062e\u062f\u0645\u062a\u0647 <a href=\"https:\/\/openai.com\/index\/learning-to-reason-with-llms\/#:~:text=test%2Dtime%20compute-,Evals,-To%20highlight%20the\" target=\"_blank\" rel=\"noreferrer noopener\">OpenAI<\/a> \u0648 <a href=\"https:\/\/github.com\/deepseek-ai\/DeepSeek-R1?tab=readme-ov-file#deepseek-r1-evaluation\" target=\"_blank\" rel=\"noreferrer noopener\">DeepSeek<\/a> \u0644\u062a\u0642\u064a\u064a\u0645 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0627\u0633\u062a\u0646\u062a\u0627\u062c \u0627\u0644\u062e\u0627\u0635\u0629 \u0628\u0647\u0645\u0627. \u064a\u0639\u062a\u0628\u0631 \u0627\u062e\u062a\u064a\u0627\u0631 GPQA-Diamond \u0643\u0645\u0639\u064a\u0627\u0631 \u062a\u0642\u064a\u064a\u0645 \u062f\u0644\u064a\u0644\u064b\u0627 \u0639\u0644\u0649 \u0635\u0639\u0648\u0628\u062a\u0647 \u0648\u0623\u0647\u0645\u064a\u062a\u0647 \u0641\u064a \u0645\u062c\u0627\u0644 \u062a\u0637\u0648\u064a\u0631 \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a.<\/p>\n<h2 id=\"heading-six\" class=\"wp-block-heading\">(6) \u0627\u0644\u0623\u062f\u0648\u0627\u062a \u0627\u0644\u0645\u0633\u062a\u062e\u062f\u0645\u0629<\/h2>\n<p class=\"wp-block-paragraph\">\u0641\u064a \u0647\u0630\u0627 \u0627\u0644\u0645\u0634\u0631\u0648\u0639\u060c \u0646\u0633\u062a\u062e\u062f\u0645 \u0628\u0634\u0643\u0644 \u0623\u0633\u0627\u0633\u064a <a href=\"http:\/\/www.ollama.com\" target=\"_blank\" rel=\"noreferrer noopener\">Ollama<\/a> \u0648 <a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https:\/\/github.com\/openai\/simple-evals\">simple-evals<\/a> \u0645\u0646 OpenAI. \u064a\u0639\u062a\u0628\u0631 Ollama \u0645\u0646\u0635\u0629 \u0642\u0648\u064a\u0629 \u0644\u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0648\u064a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631\u0629 \u0645\u062d\u0644\u064a\u064b\u0627\u060c \u0628\u064a\u0646\u0645\u0627 \u062a\u0648\u0641\u0631 simple-evals \u0625\u0637\u0627\u0631 \u0639\u0645\u0644 \u0644\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0647\u0630\u0647 \u0627\u0644\u0646\u0645\u0627\u0630\u062c.<\/p>\n<h3 class=\"wp-block-heading\">(i) \u0623\u0648\u0644\u0627\u0645\u0627 (Ollama)<\/h3>\n<p class=\"wp-block-paragraph\"><a href=\"http:\/\/ollama.com\/\" target=\"_blank\" rel=\"noreferrer noopener\"><strong>Ollama<\/strong><\/a> \u0647\u064a \u0623\u062f\u0627\u0629 \u0645\u0641\u062a\u0648\u062d\u0629 \u0627\u0644\u0645\u0635\u062f\u0631 \u062a\u0639\u0645\u0644 \u0639\u0644\u0649 \u062a\u0628\u0633\u064a\u0637 \u062a\u0634\u063a\u064a\u0644 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0627\u062a \u0627\u0644\u0643\u0628\u064a\u0631\u0629 (LLMs) \u0639\u0644\u0649 \u062c\u0647\u0627\u0632 \u0627\u0644\u0643\u0645\u0628\u064a\u0648\u062a\u0631 \u0627\u0644\u062e\u0627\u0635 \u0628\u0646\u0627 \u0623\u0648 \u0639\u0644\u0649 \u062e\u0627\u062f\u0645 \u0645\u062d\u0644\u064a. \u062a\u0639\u062a\u0628\u0631 \u0623\u0648\u0644\u0627\u0645\u0627 \u0645\u0646\u0635\u0629 \u0645\u062b\u0627\u0644\u064a\u0629 \u0644\u062a\u0634\u063a\u064a\u0644 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0645\u062d\u0644\u064a\u064b\u0627.<\/p>\n<p class=\"wp-block-paragraph\">\u062a\u0639\u0645\u0644 \u0643\u0645\u062f\u064a\u0631 \u0648\u0648\u0642\u062a \u062a\u0634\u063a\u064a\u0644\u060c \u0648\u062a\u062a\u0639\u0627\u0645\u0644 \u0645\u0639 \u0645\u0647\u0627\u0645 \u0645\u062b\u0644 \u0627\u0644\u062a\u0646\u0632\u064a\u0644\u0627\u062a \u0648\u0625\u0639\u062f\u0627\u062f \u0627\u0644\u0628\u064a\u0626\u0629. \u064a\u062a\u064a\u062d \u0630\u0644\u0643 \u0644\u0644\u0645\u0633\u062a\u062e\u062f\u0645\u064a\u0646 \u0627\u0644\u062a\u0641\u0627\u0639\u0644 \u0645\u0639 \u0647\u0630\u0647 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u062f\u0648\u0646 \u0627\u0644\u062d\u0627\u062c\u0629 \u0625\u0644\u0649 \u0627\u062a\u0635\u0627\u0644 \u062f\u0627\u0626\u0645 \u0628\u0627\u0644\u0625\u0646\u062a\u0631\u0646\u062a \u0623\u0648 \u0627\u0644\u0627\u0639\u062a\u0645\u0627\u062f \u0639\u0644\u0649 \u0627\u0644\u062e\u062f\u0645\u0627\u062a \u0627\u0644\u0633\u062d\u0627\u0628\u064a\u0629. \u062a\u0639\u062a\u0628\u0631 \u0625\u062f\u0627\u0631\u0629 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0627\u062a \u0627\u0644\u0643\u0628\u064a\u0631\u0629 (LLMs) \u0627\u0644\u0645\u062d\u0644\u064a\u0629 \u0645\u064a\u0632\u0629 \u0623\u0633\u0627\u0633\u064a\u0629 \u0641\u064a \u0623\u0648\u0644\u0627\u0645\u0627.<\/p>\n<p class=\"wp-block-paragraph\">\u062a\u062f\u0639\u0645 \u0627\u0644\u0639\u062f\u064a\u062f \u0645\u0646 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0627\u062a \u0627\u0644\u0643\u0628\u064a\u0631\u0629 \u0645\u0641\u062a\u0648\u062d\u0629 \u0627\u0644\u0645\u0635\u062f\u0631\u060c \u0628\u0645\u0627 \u0641\u064a \u0630\u0644\u0643 DeepSeek-R1\u060c \u0648\u0647\u064a \u0645\u062a\u0648\u0627\u0641\u0642\u0629 \u0639\u0628\u0631 \u0627\u0644\u0623\u0646\u0638\u0645\u0629 \u0627\u0644\u0623\u0633\u0627\u0633\u064a\u0629 \u0645\u0639 macOS \u0648 Windows \u0648 Linux. \u0628\u0627\u0644\u0625\u0636\u0627\u0641\u0629 \u0625\u0644\u0649 \u0630\u0644\u0643\u060c \u0641\u0647\u064a \u062a\u0648\u0641\u0631 \u0625\u0639\u062f\u0627\u062f\u064b\u0627 \u0645\u0628\u0627\u0634\u0631\u064b\u0627 \u0628\u0623\u0642\u0644 \u0642\u062f\u0631 \u0645\u0646 \u0627\u0644\u0636\u062c\u0629 \u0648\u0627\u0633\u062a\u062e\u062f\u0627\u0645\u064b\u0627 \u0641\u0639\u0627\u0644\u064b\u0627 \u0644\u0644\u0645\u0648\u0627\u0631\u062f. \u064a\u062a\u064a\u062d \u0644\u0643 \u0628\u0631\u0646\u0627\u0645\u062c \u0623\u0648\u0644\u0627\u0645\u0627 (Ollama) \u0627\u0644\u0627\u0633\u062a\u0641\u0627\u062f\u0629 \u0645\u0646 \u0642\u0648\u0629 \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0645\u0628\u0627\u0634\u0631\u0629 \u0639\u0644\u0649 \u062c\u0647\u0627\u0632\u0643.<\/p>\n<blockquote><p><em><strong>\u0647\u0627\u0645<\/strong>: \u062a\u0623\u0643\u062f \u0645\u0646 \u0623\u0646 \u062c\u0647\u0627\u0632\u0643 \u0627\u0644\u0645\u062d\u0644\u064a \u0644\u062f\u064a\u0647 <strong>\u0625\u0645\u0643\u0627\u0646\u064a\u0629 \u0627\u0644\u0648\u0635\u0648\u0644 \u0625\u0644\u0649 \u0648\u062d\u062f\u0629 \u0645\u0639\u0627\u0644\u062c\u0629 \u0627\u0644\u0631\u0633\u0648\u0645\u0627\u062a (GPU)<\/strong> \u0644\u0640 Ollama\u060c \u062d\u064a\u062b \u0623\u0646 \u0630\u0644\u0643 \u064a\u0633\u0631\u0639 \u0627\u0644\u0623\u062f\u0627\u0621 \u0628\u0634\u0643\u0644 \u0643\u0628\u064a\u0631 \u0648\u064a\u062c\u0639\u0644 \u0639\u0645\u0644\u064a\u0627\u062a \u0627\u0644\u0642\u064a\u0627\u0633 \u0627\u0644\u0644\u0627\u062d\u0642\u0629 \u0623\u0643\u062b\u0631 \u0643\u0641\u0627\u0621\u0629 \u0645\u0642\u0627\u0631\u0646\u0629 \u0628\u0648\u062d\u062f\u0629 \u0627\u0644\u0645\u0639\u0627\u0644\u062c\u0629 \u0627\u0644\u0645\u0631\u0643\u0632\u064a\u0629 (CPU). \u0642\u0645 \u0628\u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u0623\u0645\u0631 <code>nvidia-smi<\/code> \u0641\u064a \u0627\u0644\u0637\u0631\u0641\u064a\u0629 \u0644\u0644\u062a\u062d\u0642\u0642 \u0645\u0645\u0627 \u0625\u0630\u0627 \u062a\u0645 \u0627\u0644\u0643\u0634\u0641 \u0639\u0646 \u0648\u062d\u062f\u0629 \u0645\u0639\u0627\u0644\u062c\u0629 \u0627\u0644\u0631\u0633\u0648\u0645\u0627\u062a. \u0647\u0630\u0627 \u0627\u0644\u0625\u062c\u0631\u0627\u0621 \u064a\u0636\u0645\u0646 \u0627\u0644\u0627\u0633\u062a\u0641\u0627\u062f\u0629 \u0627\u0644\u0642\u0635\u0648\u0649 \u0645\u0646 \u0642\u062f\u0631\u0627\u062a \u0627\u0644\u062c\u0647\u0627\u0632 \u0644\u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0628\u0643\u0641\u0627\u0621\u0629 \u0639\u0627\u0644\u064a\u0629.<\/em><\/p><\/blockquote>\n<h3 class=\"wp-block-heading\">(ii) \u0645\u0643\u062a\u0628\u0629 OpenAI simple-evals \u0644\u062a\u0642\u064a\u064a\u0645 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0629<\/h3>\n<p class=\"wp-block-paragraph\">\u062a\u064f\u0639\u062f <a href=\"https:\/\/github.com\/openai\/simple-evals\" target=\"_blank\" rel=\"noreferrer noopener\"><strong>simple-evals<\/strong><\/a> \u0645\u0643\u062a\u0628\u0629 \u062e\u0641\u064a\u0641\u0629 \u0627\u0644\u0648\u0632\u0646 \u0645\u0635\u0645\u0645\u0629 \u0644\u062a\u0642\u064a\u064a\u0645 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0629 \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0645\u0646\u0647\u062c\u064a\u0629 \u0627\u0644\u062a\u0642\u064a\u064a\u0645 \u0627\u0644\u0635\u0641\u0631\u064a \u0627\u0644\u0641\u0648\u0631\u064a (zero-shot) \u0645\u0639 \u062a\u0642\u0646\u064a\u0629 &#8220;\u0633\u0644\u0633\u0644\u0629 \u0627\u0644\u0623\u0641\u0643\u0627\u0631&#8221; (chain-of-thought prompting). \u062a\u062a\u0636\u0645\u0646 \u0647\u0630\u0647 \u0627\u0644\u0645\u0643\u062a\u0628\u0629 \u0645\u0639\u0627\u064a\u064a\u0631 \u062a\u0642\u064a\u064a\u0645 \u0645\u0634\u0647\u0648\u0631\u0629 \u0645\u062b\u0644 MMLU \u0648 MATH \u0648 GPQA \u0648 MGSM \u0648 HumanEval\u060c \u0648\u062a\u0647\u062f\u0641 \u0625\u0644\u0649 \u0645\u062d\u0627\u0643\u0627\u0629 \u0633\u064a\u0646\u0627\u0631\u064a\u0648\u0647\u0627\u062a \u0627\u0644\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0627\u0644\u0648\u0627\u0642\u0639\u064a\u0629 \u0644\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0641\u064a \u0645\u0647\u0627\u0645 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0645\u0639\u0642\u062f\u0629.<\/p>\n<p class=\"wp-block-paragraph\">\u0642\u062f \u064a\u0643\u0648\u0646 \u0628\u0639\u0636\u0643\u0645 \u0639\u0644\u0649 \u062f\u0631\u0627\u064a\u0629 \u0628\u0645\u0643\u062a\u0628\u0629 \u0627\u0644\u062a\u0642\u064a\u064a\u0645 \u0627\u0644\u0623\u0634\u0647\u0631 \u0648\u0627\u0644\u0623\u0643\u062b\u0631 \u0634\u0645\u0648\u0644\u0627\u064b \u0645\u0646 OpenAI \u0648\u0627\u0644\u062a\u064a \u062a\u0633\u0645\u0649 <a href=\"https:\/\/github.com\/openai\/evals\" target=\"_blank\" rel=\"noreferrer noopener\"><strong>Evals<\/strong><\/a>\u060c \u0648\u0647\u064a \u062a\u062e\u062a\u0644\u0641 \u0639\u0646 simple-evals.<\/p>\n<p class=\"wp-block-paragraph\">\u0641\u064a \u0627\u0644\u0648\u0627\u0642\u0639\u060c \u062a\u0634\u064a\u0631 \u0635\u0641\u062d\u0629 <a href=\"https:\/\/github.com\/openai\/simple-evals?tab=readme-ov-file#background\" target=\"_blank\" rel=\"noreferrer noopener\">README<\/a> \u0627\u0644\u062e\u0627\u0635\u0629 \u0628\u0640 simple-evals \u062a\u062d\u062f\u064a\u062f\u064b\u0627 \u0625\u0644\u0649 \u0623\u0646\u0647\u0627 \u0644\u064a\u0633\u062a \u0645\u062e\u0635\u0635\u0629 \u0644\u0627\u0633\u062a\u0628\u062f\u0627\u0644 \u0645\u0643\u062a\u0628\u0629 <strong>Evals<\/strong>.<\/p>\n<p class=\"wp-block-paragraph\"><strong>\u0625\u0630\u064b\u0627\u060c \u0644\u0645\u0627\u0630\u0627 \u0646\u0633\u062a\u062e\u062f\u0645 simple-evals\u061f<\/strong><\/p>\n<p class=\"wp-block-paragraph\">\u0627\u0644\u0625\u062c\u0627\u0628\u0629 \u0627\u0644\u0628\u0633\u064a\u0637\u0629 \u0647\u064a \u0623\u0646 <strong>simple-evals<\/strong> \u062a\u0623\u062a\u064a \u0645\u0639 \u0646\u0635\u0648\u0635 \u062a\u0642\u064a\u064a\u0645 \u0645\u062f\u0645\u062c\u0629 \u0644\u0645\u0639\u0627\u064a\u064a\u0631 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u062a\u064a \u0646\u0633\u062a\u0647\u062f\u0641\u0647\u0627 (\u0645\u062b\u0644 GPQA)\u060c \u0648\u0627\u0644\u062a\u064a \u062a\u0641\u062a\u0642\u062f\u0647\u0627 \u0645\u0643\u062a\u0628\u0629 <strong>Evals<\/strong>.<\/p>\n<p class=\"wp-block-paragraph\">\u0628\u0627\u0644\u0625\u0636\u0627\u0641\u0629 \u0625\u0644\u0649 \u0630\u0644\u0643\u060c \u0644\u0645 \u0623\u062c\u062f \u0623\u064a \u0623\u062f\u0648\u0627\u062a \u0623\u0648 \u0645\u0646\u0635\u0627\u062a \u0623\u062e\u0631\u0649\u060c \u0628\u062e\u0644\u0627\u0641 simple-evals\u060c \u062a\u0648\u0641\u0631 \u0637\u0631\u064a\u0642\u0629 \u0645\u0628\u0627\u0634\u0631\u0629 \u0648\u0623\u0635\u0644\u064a\u0629 \u0628\u0644\u063a\u0629 <a target=\"_blank\" rel=\"noopener noreferrer\" title=\"Python\" href=\"https:\/\/towardsdatascience.com\/tag\/python\/\">Python<\/a> \u0644\u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u0639\u062f\u064a\u062f \u0645\u0646 \u0627\u0644\u0645\u0639\u0627\u064a\u064a\u0631 \u0627\u0644\u0631\u0626\u064a\u0633\u064a\u0629\u060c \u0645\u062b\u0644 GPQA\u060c \u062e\u0627\u0635\u0629\u064b \u0639\u0646\u062f \u0627\u0644\u0639\u0645\u0644 \u0645\u0639 Ollama.<\/p>\n<h2 id=\"heading-seven\" class=\"wp-block-heading\">(7) \u0646\u062a\u0627\u0626\u062c \u0627\u0644\u062a\u0642\u064a\u064a\u0645<\/h2>\n<p class=\"wp-block-paragraph\">\u0643\u062c\u0632\u0621 \u0645\u0646 \u0627\u0644\u062a\u0642\u064a\u064a\u0645\u060c \u0642\u0645\u062a \u0628\u0627\u062e\u062a\u064a\u0627\u0631 <strong>20 \u0633\u0624\u0627\u0644\u064b\u0627 \u0639\u0634\u0648\u0627\u0626\u064a\u064b\u0627<\/strong> \u0645\u0646 \u0645\u062c\u0645\u0648\u0639\u0629 \u0623\u0633\u0626\u0644\u0629 GPQA-Diamond \u0627\u0644\u062a\u064a \u062a\u0636\u0645 198 \u0633\u0624\u0627\u0644\u064b\u0627 \u0644\u064a\u0639\u0645\u0644 \u0639\u0644\u064a\u0647\u0627 <strong>\u0646\u0645\u0648\u0630\u062c 14B \u0627\u0644\u0645\u0642\u0637\u0631<\/strong>. \u0627\u0633\u062a\u063a\u0631\u0642 \u0627\u0644\u0623\u0645\u0631 \u0625\u062c\u0645\u0627\u0644\u064a 216 \u062f\u0642\u064a\u0642\u0629\u060c \u0623\u064a \u0645\u0627 \u064a\u0642\u0631\u0628 \u0645\u0646 11 \u062f\u0642\u064a\u0642\u0629 \u0644\u0643\u0644 \u0633\u0624\u0627\u0644.<\/p>\n<p class=\"wp-block-paragraph\">\u0643\u0627\u0646\u062a \u0627\u0644\u0646\u062a\u064a\u062c\u0629 \u0645\u062e\u064a\u0628\u0629 \u0644\u0644\u0622\u0645\u0627\u0644 \u0625\u0644\u0649 \u062d\u062f \u0645\u0627\u060c \u062d\u064a\u062b \u0633\u062c\u0644\u062a <strong>10%<\/strong> \u0641\u0642\u0637\u060c \u0648\u0647\u0648 \u0623\u0642\u0644 \u0628\u0643\u062b\u064a\u0631 \u0645\u0646 \u0627\u0644\u0646\u062a\u064a\u062c\u0629 \u0627\u0644\u0645\u0628\u0644\u063a \u0639\u0646\u0647\u0627 \u0627\u0644\u0628\u0627\u0644\u063a\u0629 73.3% \u0644\u0646\u0645\u0648\u0630\u062c DeepSeek-R1 \u0628\u062d\u062c\u0645 671B.<\/p>\n<p class=\"wp-block-paragraph\">\u0627\u0644\u0645\u0634\u0643\u0644\u0629 \u0627\u0644\u0631\u0626\u064a\u0633\u064a\u0629 \u0627\u0644\u062a\u064a \u0644\u0627\u062d\u0638\u062a\u0647\u0627 \u0647\u064a \u0623\u0646\u0647 \u062e\u0644\u0627\u0644 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u062f\u0627\u062e\u0644\u064a \u0627\u0644\u0645\u0643\u062b\u0641\u060c <strong>\u063a\u0627\u0644\u0628\u064b\u0627 \u0645\u0627 \u0641\u0634\u0644 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0625\u0645\u0627 \u0641\u064a \u0625\u0646\u062a\u0627\u062c \u0623\u064a \u0625\u062c\u0627\u0628\u0629 (\u0639\u0644\u0649 \u0633\u0628\u064a\u0644 \u0627\u0644\u0645\u062b\u0627\u0644\u060c \u0625\u0631\u062c\u0627\u0639 \u0631\u0645\u0648\u0632 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0643\u0633\u0637\u0648\u0631 \u0646\u0647\u0627\u0626\u064a\u0629 \u0644\u0644\u0625\u062e\u0631\u0627\u062c) \u0623\u0648 \u0642\u062f\u0645 \u0627\u0633\u062a\u062c\u0627\u0628\u0629 \u0644\u0645 \u062a\u062a\u0637\u0627\u0628\u0642 \u0645\u0639 \u062a\u0646\u0633\u064a\u0642 \u0627\u0644\u0627\u062e\u062a\u064a\u0627\u0631 \u0645\u0646 \u0645\u062a\u0639\u062f\u062f \u0627\u0644\u0645\u062a\u0648\u0642\u0639 (\u0639\u0644\u0649 \u0633\u0628\u064a\u0644 \u0627\u0644\u0645\u062b\u0627\u0644\u060c \u0627\u0644\u0625\u062c\u0627\u0628\u0629: \u0623).<\/strong><\/p>\n<p><img decoding=\"async\" class=\"wp-image-601945\" src=\"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2025\/05\/output_screenshot_20250420-1024x508-1.png\" alt=\"\" \/ width=\"1024\" height=\"508\"><\/p>\n<p class=\"wp-block-paragraph\">\u0643\u0645\u0627 \u0647\u0648 \u0645\u0648\u0636\u062d \u0623\u0639\u0644\u0627\u0647\u060c \u0627\u0646\u062a\u0647\u0649 \u0627\u0644\u0623\u0645\u0631 \u0628\u0627\u0644\u0639\u062f\u064a\u062f \u0645\u0646 \u0627\u0644\u0645\u062e\u0631\u062c\u0627\u062a \u0643\u0640 <code>None<\/code> \u0644\u0623\u0646 \u0645\u0646\u0637\u0642 regex \u0641\u064a simple-evals \u0644\u0645 \u064a\u062a\u0645\u0643\u0646 \u0645\u0646 \u0627\u0643\u062a\u0634\u0627\u0641 \u0646\u0645\u0637 \u0627\u0644\u0625\u062c\u0627\u0628\u0629 \u0627\u0644\u0645\u062a\u0648\u0642\u0639 \u0641\u064a \u0627\u0633\u062a\u062c\u0627\u0628\u0629 LLM.<\/p>\n<p class=\"wp-block-paragraph\">\u0641\u064a \u062d\u064a\u0646 \u0623\u0646 <a href=\"https:\/\/github.com\/kennethleungty\/DeepSeek-R1-Ollama-Simple-Evals\/blob\/main\/data\/convos_output_20250313_0105.txt\" target=\"_blank\" rel=\"noreferrer noopener\">\u0645\u0646\u0637\u0642 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0634\u0628\u064a\u0647 \u0628\u0627\u0644\u0628\u0634\u0631<\/a> \u0643\u0627\u0646 \u0645\u062b\u064a\u0631\u064b\u0627 \u0644\u0644\u0627\u0647\u062a\u0645\u0627\u0645 \u0644\u0644\u0645\u0631\u0627\u0642\u0628\u0629\u060c \u0641\u0642\u062f \u062a\u0648\u0642\u0639\u062a \u0623\u062f\u0627\u0621\u064b \u0623\u0642\u0648\u0649 \u0645\u0646 \u062d\u064a\u062b \u062f\u0642\u0629 \u0627\u0644\u0625\u062c\u0627\u0628\u0629 \u0639\u0644\u0649 \u0627\u0644\u0623\u0633\u0626\u0644\u0629.<\/p>\n<p class=\"wp-block-paragraph\">\u0644\u0642\u062f \u0631\u0623\u064a\u062a \u0623\u064a\u0636\u064b\u0627 \u0645\u0633\u062a\u062e\u062f\u0645\u064a\u0646 \u0639\u0628\u0631 \u0627\u0644\u0625\u0646\u062a\u0631\u0646\u062a \u064a\u0630\u0643\u0631\u0648\u0646 \u0623\u0646\u0647 \u062d\u062a\u0649 \u0646\u0645\u0648\u0630\u062c 32B \u0627\u0644\u0623\u0643\u0628\u0631 \u0644\u0627 \u064a\u0639\u0645\u0644 \u0628\u0634\u0643\u0644 \u062c\u064a\u062f \u0645\u062b\u0644 o1. \u0648\u0642\u062f \u0623\u062b\u0627\u0631 \u0647\u0630\u0627 \u0634\u0643\u0648\u0643\u064b\u0627 \u062d\u0648\u0644 \u0641\u0627\u0626\u062f\u0629 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0645\u0642\u0637\u0631\u0629\u060c \u062e\u0627\u0635\u0629 \u0639\u0646\u062f\u0645\u0627 \u062a\u0643\u0627\u0641\u062d \u0644\u062a\u0642\u062f\u064a\u0645 \u0625\u062c\u0627\u0628\u0627\u062a \u0635\u062d\u064a\u062d\u0629 \u0639\u0644\u0649 \u0627\u0644\u0631\u063a\u0645 \u0645\u0646 \u062a\u0648\u0644\u064a\u062f \u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0637\u0648\u064a\u0644.<\/p>\n<p class=\"wp-block-paragraph\">\u0648\u0645\u0639 \u0630\u0644\u0643\u060c \u0641\u0625\u0646 GPQA-Diamond \u0647\u0648 \u0645\u0639\u064a\u0627\u0631 \u0635\u0639\u0628 \u0644\u0644\u063a\u0627\u064a\u0629\u060c \u0644\u0630\u0644\u0643 \u0642\u062f \u062a\u0638\u0644 \u0647\u0630\u0647 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0645\u0641\u064a\u062f\u0629 \u0644\u0645\u0647\u0627\u0645 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0623\u0628\u0633\u0637. \u0643\u0645\u0627 \u0623\u0646 \u0645\u062a\u0637\u0644\u0628\u0627\u062a\u0647\u0627 \u0627\u0644\u062d\u0633\u0627\u0628\u064a\u0629 \u0627\u0644\u0623\u0642\u0644 \u062a\u062c\u0639\u0644\u0647\u0627 \u0623\u0643\u062b\u0631 \u0633\u0647\u0648\u0644\u0629.<\/p>\n<p class=\"wp-block-paragraph\">\u0639\u0644\u0627\u0648\u0629 \u0639\u0644\u0649 \u0630\u0644\u0643\u060c \u0623\u0648\u0635\u0649 \u0641\u0631\u064a\u0642 DeepSeek \u0628\u0625\u062c\u0631\u0627\u0621 \u0627\u062e\u062a\u0628\u0627\u0631\u0627\u062a \u0645\u062a\u0639\u062f\u062f\u0629 \u0648\u0645\u062a\u0648\u0633\u0637\u0629 \u0627\u0644\u0646\u062a\u0627\u0626\u062c \u0643\u062c\u0632\u0621 \u0645\u0646 \u0639\u0645\u0644\u064a\u0629 \u0627\u0644\u0642\u064a\u0627\u0633 &#8211; \u0648\u0647\u0648 \u0623\u0645\u0631 \u0623\u063a\u0641\u0644\u062a\u0647 \u0628\u0633\u0628\u0628 \u0636\u064a\u0642 \u0627\u0644\u0648\u0642\u062a.<\/p>\n<h2 id=\"heading-eight\" class=\"wp-block-heading\">(8) \u062f\u0644\u064a\u0644 \u062a\u0641\u0635\u064a\u0644\u064a \u062e\u0637\u0648\u0629 \u0628\u062e\u0637\u0648\u0629<\/h2>\n<p class=\"wp-block-paragraph\">\u062d\u062a\u0649 \u0647\u0630\u0647 \u0627\u0644\u0645\u0631\u062d\u0644\u0629\u060c \u0642\u0645\u0646\u0627 \u0628\u062a\u063a\u0637\u064a\u0629 \u0627\u0644\u0645\u0641\u0627\u0647\u064a\u0645 \u0627\u0644\u0623\u0633\u0627\u0633\u064a\u0629 \u0648\u0627\u0644\u0627\u0633\u062a\u0646\u062a\u0627\u062c\u0627\u062a \u0627\u0644\u0631\u0626\u064a\u0633\u064a\u0629.<\/p>\n<p class=\"wp-block-paragraph\">\u0625\u0630\u0627 \u0643\u0646\u062a \u0645\u0633\u062a\u0639\u062f\u064b\u0627 \u0644\u062a\u062c\u0631\u0628\u0629 \u0639\u0645\u0644\u064a\u0629 \u0648\u062a\u0642\u0646\u064a\u0629\u060c \u064a\u0642\u062f\u0645 \u0647\u0630\u0627 \u0627\u0644\u0642\u0633\u0645 \u0646\u0638\u0631\u0629 \u0645\u062a\u0639\u0645\u0642\u0629 \u0639\u0644\u0649 \u0627\u0644\u0622\u0644\u064a\u0627\u062a \u0627\u0644\u062f\u0627\u062e\u0644\u064a\u0629 \u0648\u0627\u0644\u062a\u0646\u0641\u064a\u0630 \u062e\u0637\u0648\u0629 \u0628\u062e\u0637\u0648\u0629. \u0647\u0630\u0627 \u0627\u0644\u062f\u0644\u064a\u0644 \u0627\u0644\u062a\u0642\u0646\u064a \u0627\u0644\u0639\u0645\u0644\u064a \u0633\u064a\u0648\u0641\u0631 \u0644\u0643 \u0641\u0647\u0645\u064b\u0627 \u0634\u0627\u0645\u0644\u0627\u064b \u0644\u0643\u064a\u0641\u064a\u0629 \u0639\u0645\u0644 \u0627\u0644\u0646\u0638\u0627\u0645.<\/p>\n<blockquote><p><em>\u0644\u0644\u0627\u0637\u0644\u0627\u0639 \u0639\u0644\u0649 (\u0623\u0648 \u0627\u0633\u062a\u0646\u0633\u0627\u062e) <a href=\"https:\/\/github.com\/kennethleungty\/DeepSeek-R1-Ollama-Simple-Evals\" target=\"_blank\" rel=\"noreferrer noopener\">\u0645\u0633\u062a\u0648\u062f\u0639 GitHub \u0627\u0644\u0645\u0635\u0627\u062d\u0628<\/a> \u0644\u0644\u0645\u062a\u0627\u0628\u0639\u0629. \u064a\u0645\u0643\u0646 \u0627\u0644\u0639\u062b\u0648\u0631 \u0639\u0644\u0649 \u0645\u062a\u0637\u0644\u0628\u0627\u062a \u0625\u0639\u062f\u0627\u062f \u0627\u0644\u0628\u064a\u0626\u0629 \u0627\u0644\u0627\u0641\u062a\u0631\u0627\u0636\u064a\u0629 <a href=\"https:\/\/github.com\/kennethleungty\/DeepSeek-R1-Ollama-Simple-Evals\/blob\/main\/requirements.txt\" target=\"_blank\" rel=\"noreferrer noopener\">\u0647\u0646\u0627<\/a>.<\/em><\/p><\/blockquote>\n<h3 class=\"wp-block-heading\">(i) \u0627\u0644\u0625\u0639\u062f\u0627\u062f \u0627\u0644\u0623\u0648\u0644\u064a \u2013 Ollama<\/h3>\n<p class=\"wp-block-paragraph\">\u0646\u0628\u062f\u0623 \u0628\u062a\u0646\u0632\u064a\u0644 Ollama. \u062a\u0641\u0636\u0644 \u0628\u0632\u064a\u0627\u0631\u0629 <a href=\"https:\/\/ollama.com\/download\" target=\"_blank\" rel=\"noreferrer noopener\"><br \/>\n<strong>\u0635\u0641\u062d\u0629 \u062a\u0646\u0632\u064a\u0644 Ollama<\/strong><\/a>\u060c \u0648\u0627\u062e\u062a\u0631 \u0646\u0638\u0627\u0645 \u0627\u0644\u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u062e\u0627\u0635 \u0628\u0643\u060c \u0648\u0627\u062a\u0628\u0639 \u062a\u0639\u0644\u064a\u0645\u0627\u062a \u0627\u0644\u062a\u062b\u0628\u064a\u062a \u0627\u0644\u0645\u0637\u0627\u0628\u0642\u0629.<\/p>\n<p class=\"wp-block-paragraph\">\u0628\u0645\u062c\u0631\u062f \u0627\u0643\u062a\u0645\u0627\u0644 \u0627\u0644\u062a\u062b\u0628\u064a\u062a\u060c \u0642\u0645 \u0628\u062a\u0634\u063a\u064a\u0644 Ollama \u0628\u0627\u0644\u0646\u0642\u0631 \u0627\u0644\u0645\u0632\u062f\u0648\u062c \u0639\u0644\u0649 \u062a\u0637\u0628\u064a\u0642 Ollama (\u0644\u0646\u0638\u0627\u0645\u064a \u0627\u0644\u062a\u0634\u063a\u064a\u0644 Windows \u0648 macOS) \u0623\u0648 \u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u0623\u0645\u0631 <code>ollama serve<\/code> \u0641\u064a \u0627\u0644\u0648\u062d\u062f\u0629 \u0627\u0644\u0637\u0631\u0641\u064a\u0629.<\/p>\n<h3 class=\"wp-block-heading\">(ii) \u0627\u0644\u0625\u0639\u062f\u0627\u062f \u0627\u0644\u0623\u0648\u0644\u064a \u2013 OpenAI simple-evals<\/h3>\n<p class=\"wp-block-paragraph\">\u0625\u0639\u062f\u0627\u062f simple-evals \u0641\u0631\u064a\u062f \u0646\u0633\u0628\u064a\u064b\u0627.<\/p>\n<p class=\"wp-block-paragraph\">\u0641\u064a \u062d\u064a\u0646 \u0623\u0646 simple-evals \u064a\u0642\u062f\u0645 \u0646\u0641\u0633\u0647 \u0643\u0645\u0643\u062a\u0628\u0629\u060c <strong>\u0641\u0625\u0646 \u063a\u064a\u0627\u0628 \u0645\u0644\u0641\u0627\u062a <\/strong><code>__init__.py<\/code><strong> \u0641\u064a \u0627\u0644\u0645\u0633\u062a\u0648\u062f\u0639 \u064a\u0639\u0646\u064a \u0623\u0646\u0647 \u063a\u064a\u0631 \u0645\u0647\u064a\u0643\u0644 \u0643\u062d\u0632\u0645\u0629 \u0628\u0627\u064a\u062b\u0648\u0646 \u0645\u0646\u0627\u0633\u0628\u0629<\/strong>\u060c \u0645\u0645\u0627 \u064a\u0624\u062f\u064a \u0625\u0644\u0649 \u0623\u062e\u0637\u0627\u0621 \u0627\u0644\u0627\u0633\u062a\u064a\u0631\u0627\u062f \u0628\u0639\u062f \u0627\u0633\u062a\u0646\u0633\u0627\u062e \u0627\u0644\u0645\u0633\u062a\u0648\u062f\u0639 \u0645\u062d\u0644\u064a\u064b\u0627. \u0647\u0630\u0627 \u064a\u0639\u0646\u064a \u0623\u0646\u0647 \u0644\u064a\u0633 \u062d\u0632\u0645\u0629 \u0628\u0627\u064a\u062b\u0648\u0646 \u0642\u064a\u0627\u0633\u064a\u0629 \u0628\u0627\u0644\u0645\u0639\u0646\u0649 \u0627\u0644\u0645\u062a\u0639\u0627\u0631\u0641 \u0639\u0644\u064a\u0647 \u0641\u064a \u0647\u0646\u062f\u0633\u0629 \u0627\u0644\u0628\u0631\u0645\u062c\u064a\u0627\u062a.<\/p>\n<p class=\"wp-block-paragraph\">\u0646\u0638\u0631\u064b\u0627 \u0644\u0623\u0646\u0647 \u063a\u064a\u0631 \u0645\u0646\u0634\u0648\u0631 \u0623\u064a\u0636\u064b\u0627 \u0639\u0644\u0649 PyPI \u0648\u064a\u0641\u062a\u0642\u0631 \u0625\u0644\u0649 \u0645\u0644\u0641\u0627\u062a \u0627\u0644\u062a\u0639\u0628\u0626\u0629 \u0627\u0644\u0642\u064a\u0627\u0633\u064a\u0629 \u0645\u062b\u0644 <code>setup.py<\/code> \u0623\u0648 <code>pyproject.toml<\/code>\u060c \u0641\u0644\u0627 \u064a\u0645\u0643\u0646 \u062a\u062b\u0628\u064a\u062a\u0647 \u0639\u0628\u0631 <code>pip<\/code>. \u0647\u0630\u0627 \u064a\u062c\u0639\u0644\u0647 \u062a\u062d\u062f\u064a\u064b\u0627 \u0628\u0639\u0636 \u0627\u0644\u0634\u064a\u0621 \u0644\u0644\u0645\u0637\u0648\u0631\u064a\u0646 \u0627\u0644\u062c\u062f\u062f.<\/p>\n<p class=\"wp-block-paragraph\">\u0644\u062d\u0633\u0646 \u0627\u0644\u062d\u0638\u060c \u064a\u0645\u0643\u0646\u0646\u0627 \u0627\u0633\u062a\u062e\u062f\u0627\u0645 <strong>\u0648\u062d\u062f\u0627\u062a Git \u0627\u0644\u0641\u0631\u0639\u064a\u0629<\/strong> \u0643\u062d\u0644 \u0628\u062f\u064a\u0644 \u0648\u0645\u0628\u0627\u0634\u0631. \u0647\u0630\u0647 \u0627\u0644\u0648\u062d\u062f\u0627\u062a \u062a\u0633\u0645\u062d \u0628\u062a\u0636\u0645\u064a\u0646 \u0645\u0633\u062a\u0648\u062f\u0639 Git \u062f\u0627\u062e\u0644 \u0645\u0633\u062a\u0648\u062f\u0639 \u0622\u062e\u0631\u060c \u0645\u0645\u0627 \u064a\u0633\u0647\u0644 \u0625\u062f\u0627\u0631\u0629 \u0627\u0644\u062a\u0628\u0639\u064a\u0627\u062a.<\/p>\n<p>&#8220;`html<\/p>\n<blockquote><p><em>\u062a\u0633\u0645\u062d \u0644\u0646\u0627 \u0648\u062d\u062f\u0629 Git \u0627\u0644\u0641\u0631\u0639\u064a\u0629 \u0628\u062a\u0636\u0645\u064a\u0646 \u0645\u062d\u062a\u0648\u064a\u0627\u062a \u0645\u0633\u062a\u0648\u062f\u0639 Git \u0622\u062e\u0631 \u062f\u0627\u062e\u0644 \u0645\u0634\u0631\u0648\u0639\u0646\u0627. \u062a\u0633\u062d\u0628 \u0627\u0644\u0645\u0644\u0641\u0627\u062a \u0645\u0646 \u0645\u0633\u062a\u0648\u062f\u0639 \u062e\u0627\u0631\u062c\u064a (\u0645\u062b\u0644 simple-evals)\u060c \u0648\u0644\u0643\u0646\u0647\u0627 \u062a\u062d\u062a\u0641\u0638 \u0628\u062a\u0627\u0631\u064a\u062e\u0647\u0627 \u0645\u0646\u0641\u0635\u0644\u0627\u064b.<\/em><\/p><\/blockquote>\n<p class=\"wp-block-paragraph\">\u064a\u0645\u0643\u0646\u0643 \u0627\u062e\u062a\u064a\u0627\u0631 \u0625\u062d\u062f\u0649 \u0627\u0644\u0637\u0631\u064a\u0642\u062a\u064a\u0646 (\u0623 \u0623\u0648 \u0628) \u0644\u0633\u062d\u0628 \u0645\u062d\u062a\u0648\u064a\u0627\u062a simple-evals:<\/p>\n<p class=\"wp-block-paragraph\"><strong><em>(\u0623) \u0625\u0630\u0627 \u0642\u0645\u062a \u0628\u0627\u0633\u062a\u0646\u0633\u0627\u062e \u0645\u0633\u062a\u0648\u062f\u0639 \u0645\u0634\u0631\u0648\u0639\u064a<\/em><\/strong><\/p>\n<p class=\"wp-block-paragraph\">\u064a\u062a\u0636\u0645\u0646 \u0645\u0633\u062a\u0648\u062f\u0639 \u0645\u0634\u0631\u0648\u0639\u064a \u0628\u0627\u0644\u0641\u0639\u0644 <code>simple-evals<\/code> \u0643\u0648\u062d\u062f\u0629 \u0641\u0631\u0639\u064a\u0629\u060c \u0644\u0630\u0627 \u064a\u0645\u0643\u0646\u0643 \u0641\u0642\u0637 \u062a\u0634\u063a\u064a\u0644:<\/p>\n<pre class=\"wp-block-prismatic-blocks language-bash\" tabindex=\"0\"><code class=\"language-bash\"><span class=\"token function\">git<\/span> submodule update --init --recursive<\/code><\/pre>\n<p class=\"wp-block-paragraph\"><strong><em>(\u0628) \u0625\u0630\u0627 \u0643\u0646\u062a \u062a\u0636\u064a\u0641\u0647 \u0625\u0644\u0649 \u0645\u0634\u0631\u0648\u0639 \u062a\u0645 \u0625\u0646\u0634\u0627\u0624\u0647 \u062d\u062f\u064a\u062b\u064b\u0627<\/em><\/strong><br \/>\n\u0644\u0625\u0636\u0627\u0641\u0629 simple-evals \u064a\u062f\u0648\u064a\u064b\u0627 \u0643\u0648\u062d\u062f\u0629 \u0641\u0631\u0639\u064a\u0629\u060c \u0642\u0645 \u0628\u062a\u0634\u063a\u064a\u0644 \u0647\u0630\u0627:<\/p>\n<pre class=\"wp-block-prismatic-blocks language-bash\" tabindex=\"0\"><code class=\"language-bash\"><span class=\"token function\">git<\/span> submodule add https:\/\/github.com\/openai\/simple-evals.git simple_evals<\/code><\/pre>\n<p class=\"wp-block-paragraph\"><strong>\u0645\u0644\u0627\u062d\u0638\u0629<\/strong>: \u0625\u0646 <code>simple_evals<\/code> \u0641\u064a \u0646\u0647\u0627\u064a\u0629 \u0627\u0644\u0623\u0645\u0631 (\u0645\u0639 <strong>\u0634\u0631\u0637\u0629 \u0633\u0641\u0644\u064a\u0629<\/strong>) \u0623\u0645\u0631 \u0628\u0627\u0644\u063a \u0627\u0644\u0623\u0647\u0645\u064a\u0629. \u0641\u0647\u0648 \u064a\u062d\u062f\u062f \u0627\u0633\u0645 \u0627\u0644\u0645\u062c\u0644\u062f\u060c \u0648\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0648\u0627\u0635\u0644\u0629 \u0628\u062f\u0644\u0627\u064b \u0645\u0646 \u0630\u0644\u0643 (\u0623\u064a simple<strong>\u2013<\/strong>evals) \u064a\u0645\u0643\u0646 \u0623\u0646 \u064a\u0624\u062f\u064a \u0625\u0644\u0649 \u0645\u0634\u0627\u0643\u0644 \u0641\u064a \u0627\u0644\u0627\u0633\u062a\u064a\u0631\u0627\u062f \u0644\u0627\u062d\u0642\u064b\u0627.<\/p>\n<p class=\"wp-block-paragraph\"><strong>\u0627\u0644\u062e\u0637\u0648\u0629 \u0627\u0644\u0623\u062e\u064a\u0631\u0629 (\u0644\u0643\u0644\u0627 \u0627\u0644\u0637\u0631\u064a\u0642\u062a\u064a\u0646)<\/strong><\/p>\n<p class=\"wp-block-paragraph\">\u0628\u0639\u062f \u0633\u062d\u0628 \u0645\u062d\u062a\u0648\u064a\u0627\u062a \u0627\u0644\u0645\u0633\u062a\u0648\u062f\u0639\u060c \u064a\u062c\u0628 \u0639\u0644\u064a\u0643 \u0625\u0646\u0634\u0627\u0621 \u0645\u0644\u0641 <code>__init__.py<\/code> \u0641\u0627\u0631\u063a \u0641\u064a \u0645\u062c\u0644\u062f <code>simple_evals<\/code> \u0627\u0644\u0630\u064a \u062a\u0645 \u0625\u0646\u0634\u0627\u0624\u0647 \u062d\u062f\u064a\u062b\u064b\u0627 \u062d\u062a\u0649 \u064a\u0645\u0643\u0646 \u0627\u0633\u062a\u064a\u0631\u0627\u062f\u0647 \u0643\u0648\u062d\u062f\u0629. \u064a\u0645\u0643\u0646\u0643 \u0625\u0646\u0634\u0627\u0624\u0647 \u064a\u062f\u0648\u064a\u064b\u0627\u060c \u0623\u0648 \u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0627\u0644\u0623\u0645\u0631 \u0627\u0644\u062a\u0627\u0644\u064a:<\/p>\n<pre class=\"wp-block-prismatic-blocks language-bash\" tabindex=\"0\"><code class=\"language-bash\"><span class=\"token function\">touch<\/span> simple_evals\/__init__.py<\/code><\/pre>\n<h3 class=\"wp-block-heading\">(iii) \u0633\u062d\u0628 \u0646\u0645\u0648\u0630\u062c DeepSeek-R1 \u0639\u0628\u0631 Ollama<\/h3>\n<p class=\"wp-block-paragraph\">\u0627\u0644\u062e\u0637\u0648\u0629 \u0627\u0644\u062a\u0627\u0644\u064a\u0629 \u0647\u064a \u062a\u0646\u0632\u064a\u0644 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0627\u0644\u0645\u0642\u0637\u0631 \u0627\u0644\u0630\u064a \u062a\u062e\u062a\u0627\u0631\u0647 \u0645\u062d\u0644\u064a\u064b\u0627 (\u0639\u0644\u0649 \u0633\u0628\u064a\u0644 \u0627\u0644\u0645\u062b\u0627\u0644\u060c 14B) \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0647\u0630\u0627 \u0627\u0644\u0623\u0645\u0631:<\/p>\n<p class=\"wp-block-paragraph\">\u064a\u0645\u0643\u0646 \u0627\u0644\u0639\u062b\u0648\u0631 \u0639\u0644\u0649 \u0642\u0627\u0626\u0645\u0629 \u0628\u0646\u0645\u0627\u0630\u062c DeepSeek-R1 \u0627\u0644\u0645\u062a\u0648\u0641\u0631\u0629 \u0639\u0644\u0649 Ollama <a href=\"https:\/\/ollama.com\/library\/deepseek-r1\" target=\"_blank\" rel=\"noreferrer noopener\">\u0647\u0646\u0627<\/a>. \u0644\u0644\u062d\u0635\u0648\u0644 \u0639\u0644\u0649 \u0623\u0641\u0636\u0644 \u0623\u062f\u0627\u0621\u060c \u064a\u0648\u0635\u0649 \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0623\u062d\u062f\u062b \u0625\u0635\u062f\u0627\u0631 \u0645\u0646 \u0627\u0644\u0646\u0645\u0648\u0630\u062c.<\/p>\n<pre class=\"wp-block-prismatic-blocks language-bash\" tabindex=\"0\"><code class=\"language-bash\">ollama pull deepseek-r1:14b<\/code><\/pre>\n<h3 class=\"wp-block-heading\">(\u0631\u0627\u0628\u0639\u064b\u0627) \u062a\u062d\u062f\u064a\u062f \u0627\u0644\u0625\u0639\u062f\u0627\u062f\u0627\u062a<\/h3>\n<p class=\"wp-block-paragraph\">\u0646\u0642\u0648\u0645 \u0628\u062a\u062d\u062f\u064a\u062f \u0627\u0644\u0645\u0639\u0644\u0645\u0627\u062a \u0641\u064a \u0645\u0644\u0641 YAML \u0627\u0644\u062e\u0627\u0635 \u0628\u0627\u0644\u0625\u0639\u062f\u0627\u062f\u0627\u062a\u060c \u0643\u0645\u0627 \u0647\u0648 \u0645\u0648\u0636\u062d \u0623\u062f\u0646\u0627\u0647:<\/p>\n<pre># config\/config.yaml\r\n\r\nMODEL_NAME: \"deepseek-r1:14b\" # Model name (match with Ollama model list)\r\nMODEL_TEMPERATURE: 0.6 # Set between 0.5 and 0.7 for DeepSeek-R1\r\nEVAL_BENCHMARK: \"gpqa\"\r\nGPQA_VARIANT: \"diamond\"\r\nEVAL_N_EXAMPLES: 20<\/pre>\n<p class=\"wp-block-paragraph\">\u062a\u0645 \u0636\u0628\u0637 \u062f\u0631\u062c\u0629 \u062d\u0631\u0627\u0631\u0629 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0639\u0644\u0649 <strong>0.6<\/strong> (\u0645\u0642\u0627\u0631\u0646\u0629 \u0628\u0627\u0644\u0642\u064a\u0645\u0629 \u0627\u0644\u0627\u0641\u062a\u0631\u0627\u0636\u064a\u0629 \u0627\u0644\u0646\u0645\u0648\u0630\u062c\u064a\u0629 0). \u064a\u062a\u0628\u0639 \u0647\u0630\u0627 \u062a\u0648\u0635\u064a\u0627\u062a \u0627\u0633\u062a\u062e\u062f\u0627\u0645 DeepSeek\u060c \u0627\u0644\u062a\u064a \u062a\u0642\u062a\u0631\u062d \u0646\u0637\u0627\u0642 \u062f\u0631\u062c\u0629 \u062d\u0631\u0627\u0631\u0629 \u0645\u0646 0.5 \u0625\u0644\u0649 0.7 (\u064a\u0648\u0635\u0649 \u0628\u0640 0.6) <strong>\u0644\u0645\u0646\u0639 \u0627\u0644\u062a\u0643\u0631\u0627\u0631\u0627\u062a \u0627\u0644\u0644\u0627\u0646\u0647\u0627\u0626\u064a\u0629 \u0623\u0648 \u0627\u0644\u0645\u062e\u0631\u062c\u0627\u062a \u063a\u064a\u0631 \u0627\u0644\u0645\u062a\u0645\u0627\u0633\u0643\u0629.<\/strong> \u0647\u0630\u0627 \u0627\u0644\u0625\u0639\u062f\u0627\u062f \u0636\u0631\u0648\u0631\u064a \u0644\u062a\u062d\u0633\u064a\u0646 \u062c\u0648\u062f\u0629 \u0627\u0644\u0645\u062e\u0631\u062c\u0627\u062a \u0648\u0636\u0645\u0627\u0646 \u0627\u062a\u0633\u0627\u0642\u0647\u0627.<\/p>\n<blockquote><p><em>\u0644\u0627 \u062a\u0641\u0648\u062a \u0641\u0631\u0635\u0629 \u0627\u0644\u0627\u0637\u0644\u0627\u0639 \u0639\u0644\u0649 <a href=\"https:\/\/github.com\/deepseek-ai\/DeepSeek-R1?tab=readme-ov-file#usage-recommendations\" target=\"_blank\" rel=\"noreferrer noopener\">\u062a\u0648\u0635\u064a\u0627\u062a \u0627\u0633\u062a\u062e\u062f\u0627\u0645 DeepSeek-R1 \u0627\u0644\u0641\u0631\u064a\u062f\u0629 \u0648\u0627\u0644\u0645\u062b\u064a\u0631\u0629 \u0644\u0644\u0627\u0647\u062a\u0645\u0627\u0645<\/a> \u2013 \u062e\u0627\u0635\u0629\u064b \u0641\u064a\u0645\u0627 \u064a\u062a\u0639\u0644\u0642 \u0628\u0627\u0644\u0645\u0639\u0627\u064a\u064a\u0631 \u0627\u0644\u0642\u064a\u0627\u0633\u064a\u0629 \u2013 \u0644\u0636\u0645\u0627\u0646 \u0627\u0644\u0623\u062f\u0627\u0621 \u0627\u0644\u0623\u0645\u062b\u0644 \u0639\u0646\u062f \u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0646\u0645\u0627\u0630\u062c DeepSeek-R1.<\/em><\/p><\/blockquote>\n<p class=\"wp-block-paragraph\"><code>EVAL_N_EXAMPLES<\/code> \u0647\u064a \u0627\u0644\u0645\u0639\u0644\u0645\u0629 \u0627\u0644\u0645\u0633\u062a\u062e\u062f\u0645\u0629 \u0644\u062a\u0639\u064a\u064a\u0646 \u0639\u062f\u062f \u0627\u0644\u0623\u0633\u0626\u0644\u0629 \u0645\u0646 \u0627\u0644\u0645\u062c\u0645\u0648\u0639\u0629 \u0627\u0644\u0643\u0627\u0645\u0644\u0629 \u0627\u0644\u0645\u0643\u0648\u0646\u0629 \u0645\u0646 198 \u0633\u0624\u0627\u0644\u064b\u0627 \u0648\u0627\u0644\u0645\u0633\u062a\u062e\u062f\u0645\u0629 \u0641\u064a \u0627\u0644\u062a\u0642\u064a\u064a\u0645. \u0647\u0630\u0647 \u0627\u0644\u0645\u0639\u0644\u0645\u0629 \u0636\u0631\u0648\u0631\u064a\u0629 \u0644\u0636\u0628\u0637 \u0639\u0645\u0644\u064a\u0629 \u0627\u0644\u062a\u0642\u064a\u064a\u0645 \u0648\u0641\u0642\u064b\u0627 \u0644\u0644\u0645\u0648\u0627\u0631\u062f \u0627\u0644\u0645\u062a\u0627\u062d\u0629 \u0648\u0623\u0647\u062f\u0627\u0641 \u0627\u0644\u0627\u062e\u062a\u0628\u0627\u0631 \u0627\u0644\u0645\u062d\u062f\u062f\u0629.<\/p>\n<h3 class=\"wp-block-heading\">(v) \u0625\u0639\u062f\u0627\u062f \u0643\u0648\u062f Sampler (\u0623\u062f\u0627\u0629 \u0623\u062e\u0630 \u0627\u0644\u0639\u064a\u0646\u0627\u062a)<\/h3>\n<p class=\"wp-block-paragraph\">\u0644\u062f\u0639\u0645 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0629 \u0627\u0644\u0642\u0627\u0626\u0645\u0629 \u0639\u0644\u0649 Ollama \u0636\u0645\u0646 \u0625\u0637\u0627\u0631 \u0639\u0645\u0644 simple-evals\u060c \u0646\u0642\u0648\u0645 \u0628\u0625\u0646\u0634\u0627\u0621 \u0641\u0626\u0629 \u062a\u063a\u0644\u064a\u0641 \u0645\u062e\u0635\u0635\u0629 \u0628\u0627\u0633\u0645 <code>OllamaSampler<\/code> \u0648\u062d\u0641\u0638\u0647\u0627 \u062f\u0627\u062e\u0644 <code>utils\/samplers\/ollama_sampler.py<\/code>. \u062a\u0639\u062a\u0628\u0631 \u0623\u062f\u0627\u0629 \u0623\u062e\u0630 \u0627\u0644\u0639\u064a\u0646\u0627\u062a (Sampler) \u0645\u0643\u0648\u0646\u064b\u0627 \u0623\u0633\u0627\u0633\u064a\u064b\u0627 \u0641\u064a \u0627\u062e\u062a\u0628\u0627\u0631 \u0648\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0629.<\/p>\n<pre># utils\/samplers\/ollama_sampler.py\r\n\r\nimport ollama\r\n\r\nclass OllamaSampler:\r\ndef __init__(self, model_name=None, temperature=0):\r\nself.model_name = model_name\r\nself.temperature = temperature\r\n\r\ndef __call__(self, prompt_messages):\r\nprompt_text = prompt_messages[-1][\"content\"]\r\nresponse = ollama.chat(\r\nmodel=self.model_name, \r\nmessages=[{\"role\": \"user\", \"content\": prompt_text}],\r\noptions={\"temperature\": self.temperature}\r\n)\r\nresponse_content = response[\"message\"][\"content\"]\r\n\r\nreturn response_content\r\n\r\ndef _pack_message(self, content, role):\r\nreturn {\"role\": role, \"content\": content}<\/pre>\n<p class=\"wp-block-paragraph\">\u0641\u064a \u0647\u0630\u0627 \u0627\u0644\u0633\u064a\u0627\u0642\u060c \u064a\u064f\u0642\u0635\u062f \u0628\u0640 <em>sampler<\/em> (\u0623\u062f\u0627\u0629 \u0623\u062e\u0630 \u0627\u0644\u0639\u064a\u0646\u0627\u062a) \u0641\u0626\u0629 Python \u062a\u0642\u0648\u0645 \u0628\u0625\u0646\u0634\u0627\u0621 \u0645\u062e\u0631\u062c\u0627\u062a \u0645\u0646 \u0646\u0645\u0648\u0630\u062c \u0644\u063a\u0629 \u0628\u0646\u0627\u0621\u064b \u0639\u0644\u0649 \u0645\u0637\u0627\u0644\u0628\u0629 \u0645\u0639\u064a\u0646\u0629. \u0648\u062a\u0639\u062a\u0628\u0631 \u0647\u0630\u0647 \u0627\u0644\u0623\u062f\u0627\u0629 \u062d\u0627\u0633\u0645\u0629 \u0644\u0636\u0645\u0627\u0646 \u062a\u0648\u0644\u064a\u062f \u0627\u0633\u062a\u062c\u0627\u0628\u0627\u062a \u0645\u062a\u0646\u0648\u0639\u0629 \u0648\u0645\u0645\u062b\u0644\u0629 \u0645\u0646 \u0627\u0644\u0646\u0645\u0648\u0630\u062c.<\/p>\n<p class=\"wp-block-paragraph\">\u0646\u0638\u0631\u064b\u0627 \u0644\u0623\u0646 \u0623\u062f\u0648\u0627\u062a \u0623\u062e\u0630 \u0627\u0644\u0639\u064a\u0646\u0627\u062a \u0627\u0644\u0645\u0648\u062c\u0648\u062f\u0629 \u0641\u064a simple-evals \u062a\u063a\u0637\u064a \u0641\u0642\u0637 \u0645\u0648\u0641\u0631\u064a \u062e\u062f\u0645\u0627\u062a \u0645\u062b\u0644 OpenAI \u0648 Claude\u060c \u0641\u0646\u062d\u0646 \u0628\u062d\u0627\u062c\u0629 \u0625\u0644\u0649 \u0641\u0626\u0629 sampler \u062a\u0648\u0641\u0631 \u0648\u0627\u062c\u0647\u0629 \u0645\u062a\u0648\u0627\u0641\u0642\u0629 \u0645\u0639 Ollama. \u0647\u0630\u0627 \u064a\u0636\u0645\u0646 \u0627\u0644\u062a\u0643\u0627\u0645\u0644 \u0627\u0644\u0633\u0644\u0633 \u0645\u0639 \u0625\u0637\u0627\u0631 \u0627\u0644\u062a\u0642\u064a\u064a\u0645.<\/p>\n<p class=\"wp-block-paragraph\">\u062a\u0642\u0648\u0645 <code>OllamaSampler<\/code> \u0628\u0627\u0633\u062a\u062e\u0631\u0627\u062c \u0645\u0637\u0627\u0644\u0628\u0629 \u0633\u0624\u0627\u0644 GPQA\u060c \u0648\u0625\u0631\u0633\u0627\u0644\u0647\u0627 \u0625\u0644\u0649 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0628\u062f\u0631\u062c\u0629 \u062d\u0631\u0627\u0631\u0629 \u0645\u062d\u062f\u062f\u0629\u060c \u0648\u0625\u0631\u062c\u0627\u0639 \u0627\u0633\u062a\u062c\u0627\u0628\u0629 \u0646\u0635 \u0639\u0627\u062f\u064a. \u0648\u062a\u0639\u062a\u0628\u0631 \u062f\u0631\u062c\u0629 \u0627\u0644\u062d\u0631\u0627\u0631\u0629 (Temperature) \u0645\u0639\u0644\u0645\u0629 \u0647\u0627\u0645\u0629 \u062a\u062a\u062d\u0643\u0645 \u0641\u064a \u0639\u0634\u0648\u0627\u0626\u064a\u0629 \u0627\u0644\u0645\u062e\u0631\u062c\u0627\u062a.<\/p>\n<p class=\"wp-block-paragraph\">\u064a\u062a\u0645 \u062a\u0636\u0645\u064a\u0646 \u0637\u0631\u064a\u0642\u0629 <code>_pack_message<\/code> \u0644\u0636\u0645\u0627\u0646 \u062a\u0637\u0627\u0628\u0642 \u062a\u0646\u0633\u064a\u0642 \u0627\u0644\u0625\u062e\u0631\u0627\u062c \u0645\u0639 \u0645\u0627 \u062a\u062a\u0648\u0642\u0639\u0647 \u0646\u0635\u0648\u0635 \u0627\u0644\u062a\u0642\u064a\u064a\u0645 \u0641\u064a simple-evals. \u0647\u0630\u0627 \u064a\u0636\u0645\u0646 \u0627\u0644\u062a\u0648\u0627\u0641\u0642 \u0648\u0633\u0647\u0648\u0644\u0629 \u0627\u0644\u062a\u062d\u0644\u064a\u0644.<\/p>\n<h3 class=\"wp-block-heading\">6. \u200e\u0625\u0646\u0634\u0627\u0621 \u0633\u0643\u0631\u064a\u0628\u062a \u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u062a\u0642\u064a\u064a\u0645<\/h3>\n<p class=\"wp-block-paragraph\">\u064a\u0648\u0636\u062d \u0627\u0644\u0643\u0648\u062f \u0627\u0644\u062a\u0627\u0644\u064a \u0643\u064a\u0641\u064a\u0629 \u0625\u0639\u062f\u0627\u062f \u062a\u0646\u0641\u064a\u0630 \u0627\u0644\u062a\u0642\u064a\u064a\u0645 \u0641\u064a \u0645\u0644\u0641 <code>main.py<\/code>\u060c \u0628\u0645\u0627 \u0641\u064a \u0630\u0644\u0643 \u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0627\u0644\u0641\u0626\u0629 <code>GPQAEval<\/code> \u0645\u0646 \u0645\u0643\u062a\u0628\u0629 simple-evals \u0644\u062a\u0634\u063a\u064a\u0644 \u0627\u062e\u062a\u0628\u0627\u0631 GPQA \u0627\u0644\u0645\u0639\u064a\u0627\u0631\u064a.<\/p>\n<p class=\"wp-block-paragraph\">\u0627\u0644\u062f\u0627\u0644\u0629 <code>run_eval()<\/code> \u0647\u064a \u0623\u062f\u0627\u0629 \u062a\u0634\u063a\u064a\u0644 \u062a\u0642\u064a\u064a\u0645 \u0642\u0627\u0628\u0644\u0629 \u0644\u0644\u062a\u0643\u0648\u064a\u0646 \u062a\u062e\u062a\u0628\u0631 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0627\u062a \u0627\u0644\u0643\u0628\u064a\u0631\u0629 (LLMs) \u0645\u0646 \u062e\u0644\u0627\u0644 Ollama \u0639\u0644\u0649 \u0645\u0639\u0627\u064a\u064a\u0631 \u0645\u062b\u0644 GPQA. \u0647\u0630\u0647 \u0627\u0644\u062f\u0627\u0644\u0629 \u0636\u0631\u0648\u0631\u064a\u0629 \u0644\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0628\u062f\u0642\u0629.<\/p>\n<pre># main.py\r\n\r\ndef run_eval():\r\nstart_time = time.time()\r\n\r\n# Load configuration file\r\nconfig = load_config(\"config\/config.yaml\")\r\n\r\n# Initialize Ollama sampler (wrapper around Ollama chat)\r\nollama_sampler = OllamaSampler(model_name=config[\"MODEL_NAME\"],\r\ntemperature=config[\"MODEL_TEMPERATURE\"]\r\n)\r\n\r\n# Select evaluation class to use based on EVAL_BENCHMARK\r\neval_benchmark = config[\"EVAL_BENCHMARK\"] # GPQA\r\nprint(f\"&gt;&gt;&gt; Running {eval_benchmark} evaluation\")\r\nif eval_benchmark == \"gpqa\":\r\neval_class = GPQAEval\r\neval_kwargs = {\r\n\"n_repeats\": config[\"EVAL_N_REPEATS\"], # Default 1\r\n\"num_examples\": config[\"EVAL_N_EXAMPLES\"], # Set to 20\r\n\"variant\": config[\"GPQA_VARIANT\"], # GPQA-Diamond subset\r\n}\r\nelse:\r\nraise ValueError(\r\nf\"Unknown EVAL_BENCHMARK '{eval_benchmark}'.\"\r\n)\r\n\r\n# Instantiate and run the appropriate eval\r\nevaluator = eval_class(**eval_kwargs)\r\nresults = evaluator(ollama_sampler) # Execute evaluation with sampler\r\n\r\nend_time = time.time()\r\nelapsed_seconds = end_time - start_time\r\nminutes, seconds = divmod(elapsed_seconds, 60) # Compute total time taken\r\n\r\n# Returned results is an EvalResult which includes list of SingleEvalResult and aggregated metrics\r\nprint(\"&gt;&gt;&gt;&gt; Overall Evaluation Metrics:\", results.metrics)\r\nprint(\"&gt;&gt;&gt;&gt; Score:\", results.score)\r\nprint(f\"&gt;&gt;&gt;&gt; Total Execution Time: {int(minutes)} min {seconds:.2f} sec\")\r\n\r\nif __name__ == \"__main__\":\r\n# Run GPQA evaluation execution\r\nrun_eval()<\/pre>\n<p class=\"wp-block-paragraph\">\u062a\u0642\u0648\u0645 \u0627\u0644\u062f\u0627\u0644\u0629 \u0628\u062a\u062d\u0645\u064a\u0644 \u0627\u0644\u0625\u0639\u062f\u0627\u062f\u0627\u062a \u0645\u0646 \u0645\u0644\u0641 \u0627\u0644\u062a\u0647\u064a\u0626\u0629\u060c \u0648\u0625\u0639\u062f\u0627\u062f \u0641\u0626\u0629 \u0627\u0644\u062a\u0642\u064a\u064a\u0645 \u0627\u0644\u0645\u0646\u0627\u0633\u0628\u0629 \u0645\u0646 simple-evals\u060c \u0648\u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0645\u0646 \u062e\u0644\u0627\u0644 \u0639\u0645\u0644\u064a\u0629 \u062a\u0642\u064a\u064a\u0645 \u0645\u0648\u062d\u062f\u0629. \u064a\u062a\u0645 \u062d\u0641\u0638\u0647\u0627 \u0641\u064a \u0645\u0644\u0641 <code>main.py<\/code>\u060c \u0648\u0627\u0644\u0630\u064a \u064a\u0645\u0643\u0646 \u062a\u0646\u0641\u064a\u0630\u0647 \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0627\u0644\u0623\u0645\u0631 <code>python main.py<\/code>. \u0647\u0630\u0627 \u064a\u0636\u0645\u0646 \u0639\u0645\u0644\u064a\u0629 \u062a\u0642\u064a\u064a\u0645 \u0645\u062a\u0633\u0642\u0629 \u0648\u0642\u0627\u0628\u0644\u0629 \u0644\u0644\u062a\u0643\u0631\u0627\u0631.<\/p>\n<p class=\"wp-block-paragraph\">\u0628\u0627\u062a\u0628\u0627\u0639 \u0627\u0644\u062e\u0637\u0648\u0627\u062a \u0627\u0644\u0645\u0630\u0643\u0648\u0631\u0629 \u0623\u0639\u0644\u0627\u0647\u060c \u0646\u0643\u0648\u0646 \u0642\u062f \u0646\u062c\u062d\u0646\u0627 \u0641\u064a \u0625\u0639\u062f\u0627\u062f \u0648\u062a\u0646\u0641\u064a\u0630 \u0627\u062e\u062a\u0628\u0627\u0631 GPQA-Diamond \u0627\u0644\u0645\u0639\u064a\u0627\u0631\u064a \u0639\u0644\u0649 \u0646\u0645\u0648\u0630\u062c DeepSeek-R1 distilled. \u0647\u0630\u0647 \u0627\u0644\u0639\u0645\u0644\u064a\u0629 \u062a\u0648\u0641\u0631 \u0631\u0624\u0649 \u0642\u064a\u0645\u0629 \u062d\u0648\u0644 \u0642\u062f\u0631\u0627\u062a \u0627\u0644\u0646\u0645\u0648\u0630\u062c.<\/p>\n<h2 class=\"wp-block-heading\">\u062e\u0644\u0627\u0635\u0629 \u0627\u0644\u0642\u0648\u0644<\/h2>\n<p class=\"wp-block-paragraph\">\u0641\u064a \u0647\u0630\u0647 \u0627\u0644\u0645\u0642\u0627\u0644\u0629\u060c \u0627\u0633\u062a\u0639\u0631\u0636\u0646\u0627 \u0643\u064a\u0641 \u064a\u0645\u0643\u0646\u0646\u0627 \u0627\u0644\u062c\u0645\u0639 \u0628\u064a\u0646 \u0623\u062f\u0648\u0627\u062a \u0645\u062b\u0644 Ollama \u0648 simple-evals \u0645\u0646 OpenAI \u0644\u0627\u0633\u062a\u0643\u0634\u0627\u0641 \u0648\u062a\u0642\u064a\u064a\u0645 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0645\u0646 DeepSeek-R1\u060c \u0645\u0639 \u0627\u0644\u062a\u0631\u0643\u064a\u0632 \u0639\u0644\u0649 <mark>\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631\u0629<\/mark>.<\/p>\n<p class=\"wp-block-paragraph\">\u0642\u062f \u0644\u0627 \u062a\u0636\u0627\u0647\u064a \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0627\u0644\u0623\u0635\u0644\u064a \u0630\u064a \u0627\u0644\u0640 671 \u0645\u0644\u064a\u0627\u0631 \u0645\u0639\u0644\u0645\u0629 \u0641\u064a \u0645\u0639\u0627\u064a\u064a\u0631 \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u0635\u0639\u0628\u0629 \u0645\u062b\u0644 GPQA-Diamond \u062d\u062a\u0649 \u0627\u0644\u0622\u0646. \u0648\u0645\u0639 \u0630\u0644\u0643\u060c \u0641\u0625\u0646\u0647\u0627 \u062a\u0648\u0636\u062d \u0643\u064a\u0641 \u064a\u0645\u0643\u0646 \u0644\u0644\u062a\u0642\u0637\u064a\u0631 \u062a\u0648\u0633\u064a\u0639 \u0627\u0644\u0648\u0635\u0648\u0644 \u0625\u0644\u0649 \u0642\u062f\u0631\u0627\u062a \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0627\u0644\u062e\u0627\u0635\u0629 \u0628\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631\u0629 LLM. <mark>\u062a\u062d\u0633\u064a\u0646 \u0627\u0644\u0648\u0635\u0648\u0644 \u0625\u0644\u0649 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631\u0629<\/mark> \u0647\u0648 \u0647\u062f\u0641 \u0631\u0626\u064a\u0633\u064a \u0641\u064a \u0647\u0630\u0627 \u0627\u0644\u0645\u062c\u0627\u0644.<\/p>\n<p class=\"wp-block-paragraph\">\u0639\u0644\u0649 \u0627\u0644\u0631\u063a\u0645 \u0645\u0646 \u0627\u0644\u0646\u062a\u0627\u0626\u062c \u0627\u0644\u0645\u062a\u062f\u0646\u064a\u0629 \u0641\u064a \u0627\u0644\u0645\u0647\u0627\u0645 \u0627\u0644\u0645\u0639\u0642\u062f\u0629 \u0639\u0644\u0649 \u0645\u0633\u062a\u0648\u0649 \u0627\u0644\u062f\u0643\u062a\u0648\u0631\u0627\u0647\u060c \u0642\u062f \u062a\u0638\u0644 \u0647\u0630\u0647 \u0627\u0644\u0645\u062a\u063a\u064a\u0631\u0627\u062a \u0627\u0644\u0623\u0635\u063a\u0631 \u0642\u0627\u0628\u0644\u0629 \u0644\u0644\u062a\u0637\u0628\u064a\u0642 \u0641\u064a \u0633\u064a\u0646\u0627\u0631\u064a\u0648\u0647\u0627\u062a \u0623\u0642\u0644 \u062a\u0637\u0644\u0628\u064b\u0627\u060c \u0645\u0645\u0627 \u064a\u0645\u0647\u062f \u0627\u0644\u0637\u0631\u064a\u0642 \u0644\u0644\u0646\u0634\u0631 \u0627\u0644\u0645\u062d\u0644\u064a \u0627\u0644\u0641\u0639\u0627\u0644 \u0639\u0644\u0649 \u0646\u0637\u0627\u0642 \u0623\u0648\u0633\u0639 \u0645\u0646 \u0627\u0644\u0623\u062c\u0647\u0632\u0629. \u0647\u0630\u0627 \u064a\u0633\u0627\u0647\u0645 \u0641\u064a <mark>\u0646\u0634\u0631 \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631\u0629 \u0645\u062d\u0644\u064a\u064b\u0627<\/mark> \u0628\u0643\u0641\u0627\u0621\u0629.<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"<p>\u0623\u062d\u062f\u062b \u0625\u0637\u0644\u0627\u0642 \u0644\u0646\u0645\u0648\u0630\u062c DeepSeek-R1 \u0635\u062f\u0649 \u0648\u0627\u0633\u0639\u064b\u0627 \u0641\u064a \u0645\u062c\u062a\u0645\u0639 \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0627\u0644\u0639\u0627\u0644\u0645\u064a. \u0644\u0642\u062f \u062d\u0642\u0642 \u0627\u062e\u062a\u0631\u0627\u0642\u0627\u062a \u062a\u0636\u0627\u0647\u064a \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0645\u0646 Meta \u0648 OpenAI\u060c \u0648\u062d\u0642\u0642 \u0630\u0644\u0643 \u0641\u064a \u062c\u0632\u0621 \u0635\u063a\u064a\u0631 \u0645\u0646 \u0627\u0644\u0648\u0642\u062a \u0648\u0628\u062a\u0643\u0644\u0641\u0629 \u0623\u0642\u0644 \u0628\u0643\u062b\u064a\u0631. \u0648\u0644\u0643\u0646 \u0628\u0639\u064a\u062f\u064b\u0627 \u0639\u0646 \u0627\u0644\u0639\u0646\u0627\u0648\u064a\u0646 \u0627\u0644\u0631\u0626\u064a\u0633\u064a\u0629 \u0648\u0627\u0644\u0636\u062c\u0629 \u0627\u0644\u0625\u0639\u0644\u0627\u0645\u064a\u0629 \u0639\u0628\u0631 \u0627\u0644\u0625\u0646\u062a\u0631\u0646\u062a\u060c \u0643\u064a\u0641 \u064a\u0645\u0643\u0646\u0646\u0627 \u062a\u0642\u064a\u064a\u0645 \u0642\u062f\u0631\u0627\u062a \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0641\u064a \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0645\u0639\u0627\u064a\u064a\u0631 \u0645\u0639\u062a\u0631\u0641 \u0628\u0647\u0627\u061f \u0647\u0630\u0627 \u0633\u0624\u0627\u0644 \u0645\u0647\u0645 \u0644\u062e\u0628\u0631\u0627\u0621 [&hellip;]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":3,"featured_media":3596,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[14],"tags":[],"class_list":["post-3595","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v26.6 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0646\u0645\u0627\u0630\u062c DeepSeek-R1 \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0639\u0644\u0649 GPQA \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 Ollama \u0648 simple-evals \u0645\u0646 OpenAI<\/title>\n<meta name=\"description\" content=\"\u0623\u062d\u062f\u062b \u0625\u0637\u0644\u0627\u0642 \u0644\u0646\u0645\u0648\u0630\u062c DeepSeek-R1 \u0635\u062f\u0649 \u0648\u0627\u0633\u0639\u064b\u0627 \u0641\u064a \u0645\u062c\u062a\u0645\u0639 \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0627\u0644\u0639\u0627\u0644\u0645\u064a. \u0644\u0642\u062f \u062d\u0642\u0642 \u0627\u062e\u062a\u0631\u0627\u0642\u0627\u062a \u062a\u0636\u0627\u0647\u064a \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0645\u0646 Meta \u0648 OpenAI\u060c \u0648\u062d\u0642\u0642 \u0630\u0644\u0643 \u0641\u064a \u062c\u0632\u0621 \u0635\u063a\u064a\u0631\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/\" \/>\n<meta property=\"og:locale\" content=\"ar_AR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0646\u0645\u0627\u0630\u062c DeepSeek-R1 \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0639\u0644\u0649 GPQA \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 Ollama \u0648 simple-evals \u0645\u0646 OpenAI\" \/>\n<meta property=\"og:description\" content=\"\u0623\u062d\u062f\u062b \u0625\u0637\u0644\u0627\u0642 \u0644\u0646\u0645\u0648\u0630\u062c DeepSeek-R1 \u0635\u062f\u0649 \u0648\u0627\u0633\u0639\u064b\u0627 \u0641\u064a \u0645\u062c\u062a\u0645\u0639 \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0627\u0644\u0639\u0627\u0644\u0645\u064a. \u0644\u0642\u062f \u062d\u0642\u0642 \u0627\u062e\u062a\u0631\u0627\u0642\u0627\u062a \u062a\u0636\u0627\u0647\u064a \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0645\u0646 Meta \u0648 OpenAI\u060c \u0648\u062d\u0642\u0642 \u0630\u0644\u0643 \u0641\u064a \u062c\u0632\u0621 \u0635\u063a\u064a\u0631\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/\" \/>\n<meta property=\"og:site_name\" content=\"DzTecnium - AI\" \/>\n<meta property=\"article:published_time\" content=\"2025-05-03T11:34:40+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2025-06-13T05:50:19+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2025\/05\/todd-cravens-lwACYK8ScmA-unsplash.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1920\" \/>\n\t<meta property=\"og:image:height\" content=\"1277\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Marwan Mohammed\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u0643\u064f\u062a\u0628 \u0628\u0648\u0627\u0633\u0637\u0629\" \/>\n\t<meta name=\"twitter:data1\" content=\"Marwan Mohammed\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u0648\u0642\u062a \u0627\u0644\u0642\u0631\u0627\u0621\u0629 \u0627\u0644\u0645\u064f\u0642\u062f\u0651\u0631\" \/>\n\t<meta name=\"twitter:data2\" content=\"20 \u062f\u0642\u064a\u0642\u0629\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/#article\",\"isPartOf\":{\"@id\":\"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/\"},\"author\":{\"name\":\"Marwan Mohammed\",\"@id\":\"https:\/\/www.dztecs.com\/ai\/#\/schema\/person\/a4f2e05888a5694aa5d25232345f2d23\"},\"headline\":\"\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0646\u0645\u0627\u0630\u062c DeepSeek-R1 \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0639\u0644\u0649 GPQA \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 Ollama \u0648 simple-evals \u0645\u0646 OpenAI\",\"datePublished\":\"2025-05-03T11:34:40+00:00\",\"dateModified\":\"2025-06-13T05:50:19+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/\"},\"wordCount\":247,\"publisher\":{\"@id\":\"https:\/\/www.dztecs.com\/ai\/#organization\"},\"image\":{\"@id\":\"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2025\/05\/todd-cravens-lwACYK8ScmA-unsplash.jpg\",\"articleSection\":[\"\u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a\"],\"inLanguage\":\"ar\"},{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/\",\"url\":\"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/\",\"name\":\"\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0646\u0645\u0627\u0630\u062c DeepSeek-R1 \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0639\u0644\u0649 GPQA \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 Ollama \u0648 simple-evals \u0645\u0646 OpenAI\",\"isPartOf\":{\"@id\":\"https:\/\/www.dztecs.com\/ai\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/#primaryimage\"},\"image\":{\"@id\":\"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2025\/05\/todd-cravens-lwACYK8ScmA-unsplash.jpg\",\"datePublished\":\"2025-05-03T11:34:40+00:00\",\"dateModified\":\"2025-06-13T05:50:19+00:00\",\"description\":\"\u0623\u062d\u062f\u062b \u0625\u0637\u0644\u0627\u0642 \u0644\u0646\u0645\u0648\u0630\u062c DeepSeek-R1 \u0635\u062f\u0649 \u0648\u0627\u0633\u0639\u064b\u0627 \u0641\u064a \u0645\u062c\u062a\u0645\u0639 \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0627\u0644\u0639\u0627\u0644\u0645\u064a. \u0644\u0642\u062f \u062d\u0642\u0642 \u0627\u062e\u062a\u0631\u0627\u0642\u0627\u062a \u062a\u0636\u0627\u0647\u064a \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0645\u0646 Meta \u0648 OpenAI\u060c \u0648\u062d\u0642\u0642 \u0630\u0644\u0643 \u0641\u064a \u062c\u0632\u0621 \u0635\u063a\u064a\u0631\",\"breadcrumb\":{\"@id\":\"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/#breadcrumb\"},\"inLanguage\":\"ar\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"ar\",\"@id\":\"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/#primaryimage\",\"url\":\"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2025\/05\/todd-cravens-lwACYK8ScmA-unsplash.jpg\",\"contentUrl\":\"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2025\/05\/todd-cravens-lwACYK8ScmA-unsplash.jpg\",\"width\":1920,\"height\":1277},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u0627\u0644\u0631\u0626\u064a\u0633\u064a\u0629\",\"item\":\"https:\/\/www.dztecs.com\/ai\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0646\u0645\u0627\u0630\u062c DeepSeek-R1 \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0639\u0644\u0649 GPQA \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 Ollama \u0648 simple-evals \u0645\u0646 OpenAI\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.dztecs.com\/ai\/#website\",\"url\":\"https:\/\/www.dztecs.com\/ai\/\",\"name\":\"DzTecnium - AI\",\"description\":\"\u0627\u0628\u062a\u0643\u0627\u0631\u0627\u062a \u0648\u062a\u0642\u0646\u064a\u0627\u062a \u0645\u062a\u0642\u062f\u0645\u0629 \u0644\u062a\u0633\u0647\u064a\u0644 \u062d\u064a\u0627\u062a\u0643\",\"publisher\":{\"@id\":\"https:\/\/www.dztecs.com\/ai\/#organization\"},\"alternateName\":\"DzT - AI\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.dztecs.com\/ai\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"ar\"},{\"@type\":\"Organization\",\"@id\":\"https:\/\/www.dztecs.com\/ai\/#organization\",\"name\":\"DzTecnium - AI\",\"alternateName\":\"DzT - AI\",\"url\":\"https:\/\/www.dztecs.com\/ai\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"ar\",\"@id\":\"https:\/\/www.dztecs.com\/ai\/#\/schema\/logo\/image\/\",\"url\":\"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2024\/11\/ai-520px-.png\",\"contentUrl\":\"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2024\/11\/ai-520px-.png\",\"width\":805,\"height\":235,\"caption\":\"DzTecnium - AI\"},\"image\":{\"@id\":\"https:\/\/www.dztecs.com\/ai\/#\/schema\/logo\/image\/\"}},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.dztecs.com\/ai\/#\/schema\/person\/a4f2e05888a5694aa5d25232345f2d23\",\"name\":\"Marwan Mohammed\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"ar\",\"@id\":\"https:\/\/www.dztecs.com\/ai\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/secure.gravatar.com\/avatar\/e992ef113ad8c5d3afa0639a825cfa8e333d9a8fa9ceea63862437af88b49451?s=96&d=monsterid&r=g\",\"contentUrl\":\"https:\/\/secure.gravatar.com\/avatar\/e992ef113ad8c5d3afa0639a825cfa8e333d9a8fa9ceea63862437af88b49451?s=96&d=monsterid&r=g\",\"caption\":\"Marwan Mohammed\"}}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0646\u0645\u0627\u0630\u062c DeepSeek-R1 \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0639\u0644\u0649 GPQA \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 Ollama \u0648 simple-evals \u0645\u0646 OpenAI","description":"\u0623\u062d\u062f\u062b \u0625\u0637\u0644\u0627\u0642 \u0644\u0646\u0645\u0648\u0630\u062c DeepSeek-R1 \u0635\u062f\u0649 \u0648\u0627\u0633\u0639\u064b\u0627 \u0641\u064a \u0645\u062c\u062a\u0645\u0639 \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0627\u0644\u0639\u0627\u0644\u0645\u064a. \u0644\u0642\u062f \u062d\u0642\u0642 \u0627\u062e\u062a\u0631\u0627\u0642\u0627\u062a \u062a\u0636\u0627\u0647\u064a \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0645\u0646 Meta \u0648 OpenAI\u060c \u0648\u062d\u0642\u0642 \u0630\u0644\u0643 \u0641\u064a \u062c\u0632\u0621 \u0635\u063a\u064a\u0631","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/","og_locale":"ar_AR","og_type":"article","og_title":"\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0646\u0645\u0627\u0630\u062c DeepSeek-R1 \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0639\u0644\u0649 GPQA \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 Ollama \u0648 simple-evals \u0645\u0646 OpenAI","og_description":"\u0623\u062d\u062f\u062b \u0625\u0637\u0644\u0627\u0642 \u0644\u0646\u0645\u0648\u0630\u062c DeepSeek-R1 \u0635\u062f\u0649 \u0648\u0627\u0633\u0639\u064b\u0627 \u0641\u064a \u0645\u062c\u062a\u0645\u0639 \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0627\u0644\u0639\u0627\u0644\u0645\u064a. \u0644\u0642\u062f \u062d\u0642\u0642 \u0627\u062e\u062a\u0631\u0627\u0642\u0627\u062a \u062a\u0636\u0627\u0647\u064a \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0645\u0646 Meta \u0648 OpenAI\u060c \u0648\u062d\u0642\u0642 \u0630\u0644\u0643 \u0641\u064a \u062c\u0632\u0621 \u0635\u063a\u064a\u0631","og_url":"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/","og_site_name":"DzTecnium - AI","article_published_time":"2025-05-03T11:34:40+00:00","article_modified_time":"2025-06-13T05:50:19+00:00","og_image":[{"width":1920,"height":1277,"url":"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2025\/05\/todd-cravens-lwACYK8ScmA-unsplash.jpg","type":"image\/jpeg"}],"author":"Marwan Mohammed","twitter_card":"summary_large_image","twitter_misc":{"\u0643\u064f\u062a\u0628 \u0628\u0648\u0627\u0633\u0637\u0629":"Marwan Mohammed","\u0648\u0642\u062a \u0627\u0644\u0642\u0631\u0627\u0621\u0629 \u0627\u0644\u0645\u064f\u0642\u062f\u0651\u0631":"20 \u062f\u0642\u064a\u0642\u0629"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/#article","isPartOf":{"@id":"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/"},"author":{"name":"Marwan Mohammed","@id":"https:\/\/www.dztecs.com\/ai\/#\/schema\/person\/a4f2e05888a5694aa5d25232345f2d23"},"headline":"\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0646\u0645\u0627\u0630\u062c DeepSeek-R1 \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0639\u0644\u0649 GPQA \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 Ollama \u0648 simple-evals \u0645\u0646 OpenAI","datePublished":"2025-05-03T11:34:40+00:00","dateModified":"2025-06-13T05:50:19+00:00","mainEntityOfPage":{"@id":"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/"},"wordCount":247,"publisher":{"@id":"https:\/\/www.dztecs.com\/ai\/#organization"},"image":{"@id":"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/#primaryimage"},"thumbnailUrl":"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2025\/05\/todd-cravens-lwACYK8ScmA-unsplash.jpg","articleSection":["\u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a"],"inLanguage":"ar"},{"@type":"WebPage","@id":"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/","url":"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/","name":"\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0646\u0645\u0627\u0630\u062c DeepSeek-R1 \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0639\u0644\u0649 GPQA \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 Ollama \u0648 simple-evals \u0645\u0646 OpenAI","isPartOf":{"@id":"https:\/\/www.dztecs.com\/ai\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/#primaryimage"},"image":{"@id":"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/#primaryimage"},"thumbnailUrl":"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2025\/05\/todd-cravens-lwACYK8ScmA-unsplash.jpg","datePublished":"2025-05-03T11:34:40+00:00","dateModified":"2025-06-13T05:50:19+00:00","description":"\u0623\u062d\u062f\u062b \u0625\u0637\u0644\u0627\u0642 \u0644\u0646\u0645\u0648\u0630\u062c DeepSeek-R1 \u0635\u062f\u0649 \u0648\u0627\u0633\u0639\u064b\u0627 \u0641\u064a \u0645\u062c\u062a\u0645\u0639 \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0627\u0644\u0639\u0627\u0644\u0645\u064a. \u0644\u0642\u062f \u062d\u0642\u0642 \u0627\u062e\u062a\u0631\u0627\u0642\u0627\u062a \u062a\u0636\u0627\u0647\u064a \u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u0645\u0646 Meta \u0648 OpenAI\u060c \u0648\u062d\u0642\u0642 \u0630\u0644\u0643 \u0641\u064a \u062c\u0632\u0621 \u0635\u063a\u064a\u0631","breadcrumb":{"@id":"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/#breadcrumb"},"inLanguage":"ar","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/"]}]},{"@type":"ImageObject","inLanguage":"ar","@id":"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/#primaryimage","url":"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2025\/05\/todd-cravens-lwACYK8ScmA-unsplash.jpg","contentUrl":"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2025\/05\/todd-cravens-lwACYK8ScmA-unsplash.jpg","width":1920,"height":1277},{"@type":"BreadcrumbList","@id":"https:\/\/www.dztecs.com\/ai\/how-to-benchmark-deepseek-r1-distilled-models-on-gpqa-using-ollama-and-openais-simple-evals\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u0627\u0644\u0631\u0626\u064a\u0633\u064a\u0629","item":"https:\/\/www.dztecs.com\/ai\/"},{"@type":"ListItem","position":2,"name":"\u062a\u0642\u064a\u064a\u0645 \u0623\u062f\u0627\u0621 \u0646\u0645\u0627\u0630\u062c DeepSeek-R1 \u0627\u0644\u0645\u0642\u0637\u0631\u0629 \u0639\u0644\u0649 GPQA \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 Ollama \u0648 simple-evals \u0645\u0646 OpenAI"}]},{"@type":"WebSite","@id":"https:\/\/www.dztecs.com\/ai\/#website","url":"https:\/\/www.dztecs.com\/ai\/","name":"DzTecnium - AI","description":"\u0627\u0628\u062a\u0643\u0627\u0631\u0627\u062a \u0648\u062a\u0642\u0646\u064a\u0627\u062a \u0645\u062a\u0642\u062f\u0645\u0629 \u0644\u062a\u0633\u0647\u064a\u0644 \u062d\u064a\u0627\u062a\u0643","publisher":{"@id":"https:\/\/www.dztecs.com\/ai\/#organization"},"alternateName":"DzT - AI","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.dztecs.com\/ai\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"ar"},{"@type":"Organization","@id":"https:\/\/www.dztecs.com\/ai\/#organization","name":"DzTecnium - AI","alternateName":"DzT - AI","url":"https:\/\/www.dztecs.com\/ai\/","logo":{"@type":"ImageObject","inLanguage":"ar","@id":"https:\/\/www.dztecs.com\/ai\/#\/schema\/logo\/image\/","url":"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2024\/11\/ai-520px-.png","contentUrl":"https:\/\/www.dztecs.com\/ai\/wp-content\/uploads\/sites\/2\/2024\/11\/ai-520px-.png","width":805,"height":235,"caption":"DzTecnium - AI"},"image":{"@id":"https:\/\/www.dztecs.com\/ai\/#\/schema\/logo\/image\/"}},{"@type":"Person","@id":"https:\/\/www.dztecs.com\/ai\/#\/schema\/person\/a4f2e05888a5694aa5d25232345f2d23","name":"Marwan Mohammed","image":{"@type":"ImageObject","inLanguage":"ar","@id":"https:\/\/www.dztecs.com\/ai\/#\/schema\/person\/image\/","url":"https:\/\/secure.gravatar.com\/avatar\/e992ef113ad8c5d3afa0639a825cfa8e333d9a8fa9ceea63862437af88b49451?s=96&d=monsterid&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/e992ef113ad8c5d3afa0639a825cfa8e333d9a8fa9ceea63862437af88b49451?s=96&d=monsterid&r=g","caption":"Marwan Mohammed"}}]}},"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/www.dztecs.com\/ai\/wp-json\/wp\/v2\/posts\/3595","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.dztecs.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.dztecs.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.dztecs.com\/ai\/wp-json\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/www.dztecs.com\/ai\/wp-json\/wp\/v2\/comments?post=3595"}],"version-history":[{"count":0,"href":"https:\/\/www.dztecs.com\/ai\/wp-json\/wp\/v2\/posts\/3595\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.dztecs.com\/ai\/wp-json\/wp\/v2\/media\/3596"}],"wp:attachment":[{"href":"https:\/\/www.dztecs.com\/ai\/wp-json\/wp\/v2\/media?parent=3595"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.dztecs.com\/ai\/wp-json\/wp\/v2\/categories?post=3595"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.dztecs.com\/ai\/wp-json\/wp\/v2\/tags?post=3595"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}