{"id":27835,"date":"2025-01-26T08:25:29","date_gmt":"2025-01-26T00:25:29","guid":{"rendered":"https:\/\/www.1ai.net\/?p=27835"},"modified":"2025-01-26T08:25:29","modified_gmt":"2025-01-26T00:25:29","slug":"%e4%b8%ad%e5%9b%bd%e7%94%b5%e4%bf%a1%e5%8f%91%e5%b8%83%e5%a4%8d%e6%9d%82%e6%8e%a8%e7%90%86%e5%a4%a7%e6%a8%a1%e5%9e%8bteleai-t1-preview%ef%bc%9a%e8%83%bd%e8%a7%a3%e3%80%8a%e4%b9%9d","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/27835.html","title":{"rendered":"China Telecom releases TeleAI-t1-preview, a large model of complex reasoning that can solve the questions of the Nine Chapters of the Mathematical Art."},"content":{"rendered":"<p>1AI from<a href=\"https:\/\/www.1ai.net\/en\/tag\/%e4%b8%ad%e5%9b%bd%e7%94%b5%e4%bf%a1\" title=\"[View articles tagged with [China Telecom]]\" target=\"_blank\" >China Telecom<\/a>The Artificial Intelligence Institute has learned that its \"<a href=\"https:\/\/www.1ai.net\/en\/tag\/%e5%a4%8d%e6%9d%82%e6%8e%a8%e7%90%86%e5%a4%a7%e6%a8%a1%e5%9e%8b\" title=\"[Sees articles with labels of complex reasoning]\" target=\"_blank\" >Large models of complex reasoning<\/a>\u201d<a href=\"https:\/\/www.1ai.net\/en\/tag\/teleai\" title=\"[View articles tagged with [TeleAI]]\" target=\"_blank\" >TeleAI<\/a>TeleAI-t1-preview is now officially released and will soon be launched on the Sky AI open platform. TeleAI-t1-preview uses a reinforcement learning training methodology, which dramatically improves the accuracy of the model in complex problems such as logical reasoning and mathematical derivation by introducing the paradigms of thinking such as exploration and reflection.<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-27836\" title=\"ddf974a6j00sqo6hb0048d000u000g7p\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2025\/01\/ddf974a6j00sqo6hb0048d000u000g7p.jpg\" alt=\"ddf974a6j00sqo6hb0048d000u000g7p\" width=\"1080\" height=\"583\" \/><\/p>\n<p>Officially, TeleAI-t1-preview has been recognized as the best in the U.S. math competitions AIME 2024 and MATH500 math benchmarks.<strong>\u00a0<\/strong><strong>60 and 93.8 points<\/strong>achievements.<strong>outperform by a wide margin<\/strong><strong>\u00a0<\/strong>Benchmarking models such as OpenAI o1-preview and GPT-4o. On the graduate-level quiz test GPQA Diamond, TeleAI-t1-preview scored<strong>Over GPT-4o<\/strong>It's also on par with the performance level of the Claude 3.5 Sonnet.<\/p>\n<p>The evaluation showed that when a question from the Nine Chapters of the Mathematical Art was given to TeleAI-t1-preview, it was able to comprehend and simplify the Chinese text before converting it into modern Chinese, and then giving the mathematical derivation and answer.<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-27837\" title=\"0d5171efj00sqo6gu0069d000hs00dcm\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2025\/01\/0d5171efj00sqo6gu0069d000hs00dcm.jpg\" alt=\"0d5171efj00sqo6gu0069d000hs00dcm\" width=\"640\" height=\"480\" \/><\/p>\n<p>It is reported that in this process, TeleAI-t1-preview can combine figurative thinking with abstract thinking to visualize the scenarios involved and assist in understanding the topic. Not only that, it is also capable of rigorously converting ancient and modern units.<\/p>\n<p>TeleAI introduces innovative training strategies to ensure that the reasoning process is accurate and effective.<\/p>\n<ul>\n<li><strong>Data preparation phase:<\/strong>A high-quality reasoning dataset with a mathematical core, complemented by a multidisciplinary approach, was collected and constructed to ensure that the model can be adapted to different types of reasoning tasks.<\/li>\n<li><strong>Judge Model:<\/strong>A Judge Model was trained specifically to analyze and assess the correctness of the model's long thought links, providing guidance for model reflection and error correction.<\/li>\n<li><strong>SFT (supervised fine tuning) phase:<\/strong>MCTS (Monte Carlo Tree Search) is used to construct high-quality long reasoning data, combining the accuracy rate and solution length of each step to select the optimal complete path, which ensures the accuracy of the reasoning answer while effectively lengthening the chain of thought to obtain a more fine-grained reasoning process. At the same time, the Judge Model is used to analyze the paths with low correctness in the reasoning process, and guide the model to reflect and correct the wrong reasoning steps, so as to construct high-quality thought chain data for SFT training.<\/li>\n<li><strong>Intensive learning phase:<\/strong>The Rule-based Reward Model was additionally constructed to provide sufficiently accurate reward signals to further enhance the model's logical reasoning ability through online reinforcement learning algorithms.<\/li>\n<\/ul>","protected":false},"excerpt":{"rendered":"<p>1AI learned from China Telecom's Artificial Intelligence Research Institute that its \"complex reasoning model\" TeleAI-t1-preview has been formally released and will soon be launched on the Tianyi AI open platform. TeleAI-t1-preview uses a reinforcement learning training method to greatly improve the accuracy of the model for complex problems such as logical reasoning and mathematical derivation by introducing a paradigm of exploration and reflection. TeleAI-t1-preview uses reinforcement learning training methods, and by introducing thinking paradigms such as exploration and reflection, it significantly improves the accuracy of the model in complex problems such as logical reasoning and mathematical deduction. Officially, TeleAI-t1-preview has significantly surpassed benchmark models such as OpenAI o1-preview and GPT-4o in two math benchmarks, AIME 2024 and MATH500, with scores of 60 and 93.8 respectively. At the graduate level<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[146],"tags":[4827,3419,5642],"collection":[],"class_list":["post-27835","post","type-post","status-publish","format-standard","hentry","category-news","tag-teleai","tag-3419","tag-5642"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/27835","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=27835"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/27835\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=27835"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=27835"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=27835"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=27835"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}