{"id":222922,"date":"2025-10-05T00:05:40","date_gmt":"2025-10-05T05:05:40","guid":{"rendered":"https:\/\/lifeboat.com\/blog\/2025\/10\/new-technique-auto-selects-training-examples-to-speed-up-fine-tuning"},"modified":"2025-10-05T00:05:40","modified_gmt":"2025-10-05T05:05:40","slug":"new-technique-auto-selects-training-examples-to-speed-up-fine-tuning","status":"publish","type":"post","link":"https:\/\/lifeboat.com\/blog\/2025\/10\/new-technique-auto-selects-training-examples-to-speed-up-fine-tuning","title":{"rendered":"New Technique Auto-Selects Training Examples to Speed Up Fine-Tuning"},"content":{"rendered":"

<\/a><\/p>\n

Fine-tuning large language models via reinforcement learning is computationally expensive, but researchers found a way to streamline the process.<\/p>\n

What\u2019s new:<\/strong> Qinsi Wang and colleagues at UC Berkeley and Duke University developed GAIN-RL<\/a>, a method that accelerates reinforcement learning fine-tuning by selecting training examples automatically based on the model\u2019s own internal signals, specifically the angles between vector representations of tokens. The code is available<\/a> on GitHub.<\/p>\n

Key insight:<\/strong> The cosine similarity between a model\u2019s vector representations of input tokens governs the magnitude of gradient updates during training. Specifically, the sum of those similarities that enter a model\u2019s classification layer, called the angle concentration, governs the magnitude of gradient updates. Examples with higher angle concentration produce larger gradient updates. The magnitude of a gradient update in turn determines the effectiveness of a given training example: The larger the update, the more the model learns. Prioritizing the most-effective examples before transitioning to less-effective ones enhances training efficiency while adding little preprocessing overhead.<\/p>\n","protected":false},"excerpt":{"rendered":"

Fine-tuning large language models via reinforcement learning is computationally expensive, but researchers found a way to streamline the process. What\u2019s new: Qinsi Wang and colleagues at UC Berkeley and Duke University developed GAIN-RL, a method that accelerates reinforcement learning fine-tuning by selecting training examples automatically based on the model\u2019s own internal signals, specifically the angles [\u2026]<\/p>\n","protected":false},"author":662,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6,1491],"tags":[],"class_list":["post-222922","post","type-post","status-publish","format-standard","hentry","category-robotics-ai","category-transportation"],"_links":{"self":[{"href":"https:\/\/lifeboat.com\/blog\/wp-json\/wp\/v2\/posts\/222922","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/lifeboat.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/lifeboat.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/lifeboat.com\/blog\/wp-json\/wp\/v2\/users\/662"}],"replies":[{"embeddable":true,"href":"https:\/\/lifeboat.com\/blog\/wp-json\/wp\/v2\/comments?post=222922"}],"version-history":[{"count":0,"href":"https:\/\/lifeboat.com\/blog\/wp-json\/wp\/v2\/posts\/222922\/revisions"}],"wp:attachment":[{"href":"https:\/\/lifeboat.com\/blog\/wp-json\/wp\/v2\/media?parent=222922"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/lifeboat.com\/blog\/wp-json\/wp\/v2\/categories?post=222922"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/lifeboat.com\/blog\/wp-json\/wp\/v2\/tags?post=222922"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}