English(EN) Outside the Purview of the Reward: Capability Drift in Multilingual Post-Training

研究发现AI模型在多语言训练中出现能力漂移

作者 PulseAugur 编辑部 · [1 个来源] · 2026-11-14 05:00

一篇新的arXiv论文探讨了AI模型多语言后训练中的能力漂移问题，特别考察了一个GRPO模型的两个分支。一个分支在数学任务上取得了显著进步，但在印度语言理解方面却明显下降；而另一个分支则恢复了印度语言分数，但牺牲了其他能力。研究强调了基于奖励的训练如何会无意中削弱必要技能，并建议在解读基准分数时，应同时考虑它们所奖励的响应格式。 AI

影响凸显了AI模型训练中可能出现的权衡，这些权衡可能会影响模型在不同语言和任务上的表现。

排序理由发布在arXiv上的学术论文，详细介绍了AI模型的能力。[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; 发布在arXiv上的学术论文，详细介绍了AI模型的能力。[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, model release; 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) · Sutanshu Saksena Raj, Purusharth Malik, Pasunuti Prasanjith, Daksh Maan, Parijat Chatterjee, Avinash Benki · 2026-11-14 05:00

奖励范围之外：多语言训练后能力漂移

arXiv:2611.15717v1 Announce Type: cross Abstract: Post-training can improve rewarded behavior while weakening capabilities users still need. We study two GRPO branches from a shared supervised checkpoint in a multilingual mixture-of-experts model. The released branch gains 36.07 …

本文如何被排名

报道来源 [1]

奖励范围之外：多语言训练后能力漂移

相关实体

相关话题