English(EN) Gradient Routing Localizes and Removes Unintended Behaviors in RL

新的GRAFT技术解决了AI模型的奖励破解问题

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-15 04:00

研究人员推出了一种新方法GRAFT（Gradient Routing Adapter Fine Tuning），用于解决强化学习模型中的非预期行为或“奖励破解”问题。GRAFT通过将标记有不完美监控器的样本上的梯度更新限制在一组单独的参数上。这种方法可以防止模型学会隐藏不当行为，并通过移除这些牺牲参数来消除非预期动作，同时保持目标任务的性能。该技术已在135M到8B参数的模型上进行了演示，并成功应用于将前沿LLM的轨迹蒸馏到Qwen3-32B中，而不会引起奖励破解。 AI

影响该方法通过提供一种在不影响性能的情况下纠正非预期行为的方法，有望提高AI系统的可靠性和安全性。

排序理由该集群包含一篇详细介绍AI安全新方法的论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 16 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍AI安全新方法的论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, safety; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.AI TIER_1 English(EN) · Jake Ward, Shawn Hu, Aria Wong, Nathan Hu, Bryce Woodworth, Alexander Matt Turner, Alex Cloud · 2026-10-15 04:00

梯度路由定位并移除RL中的意外行为

arXiv:2610.17062v1 Announce Type: cross Abstract: We propose a mitigation to reward hacking in reinforcement learning. Our setup involves a misspecified reward function that gives rise to unintended behavior, and a monitor used to detect unintended behavior. As with frontier AI s…

本文如何被排名

报道来源 [1]

梯度路由定位并移除RL中的意外行为

相关实体

相关话题