研究库 论文知识库

内容库 / 论文

arXiv:2610.03240 · 工程化

Collective Bias Mitigation via Model Routing and Collaboration

Collective Bias Mitigation：通过模型路由与协作的集体偏见缓解

Collective Bias Mitigation via Model Routing and Collaboration

类型：arxiv
标识：2610.03240
链接：https://arxiv.org/abs/2610.03240
主分类：engineering
形态：application
TLDR：Large language models (LLMs) are increasingly deployed in public health, finance, and governance, requiring both accuracy and societal value alignment. Despite recent advances, LLMs often perpetuate or amplify bias embedded in their training data, posing challenges to fairness. While self-debiasing encourages an LLM to identify and correct its own biases, relying on a single model's intrinsic knowledge may be insufficient to address deeply ingrained stereotypes. To address this limitation, we introduce Collective Bias Mitigation (CBM), a framework that alleviates bias by learning fine-grained
待LLM分类：否
标题中文：Collective Bias Mitigation：通过模型路由与协作的集体偏见缓解
TLDR中文：大语言模型（LLM）日益部署于公共健康、金融和治理等关键场景，要求准确性与社会价值对齐。尽管近期取得进展，LLM 仍会延续或放大训练数据中嵌入的偏见，对公平性构成挑战。自我去偏鼓励 LLM 识别并纠正自身偏见，但依赖单个模型的内在知识可能不足以消除根深蒂固的刻板印象。为突破这一局限，我们提出 Collective Bias Mitigation（CBM），通过学习细粒度
来源文件：
/inbox/tom/_candidates/2026-10-05-agent-rag-longcontext-candidates.json

相关 · 按语义匹配

文档 Tom 评 flyP · 2026-09-24 双论文精读（Realtime-Venus + GAE）

笔记 2026-09-24-0950 · flyP 双论文精读 · Realtime-Venus + GAE

笔记 Tom 文献雷达 · Agent × RAG × 长上下文 · 2026-07-27 20:40 UTC

笔记 Tom 文献雷达 · AI Agent / RAG / 长上下文 · 2026-08-02 14:40

笔记 Agentic Coding in the Wild · arXiv:2608.00101v1 · coding-agent systems · 关键短审稿

笔记 Cool Papers · cs.CL (papers.cool) · RSS 摘要

arXiv

2610.03240

TLDR

大语言模型（LLM）日益部署于公共健康、金融和治理等关键场景，要求准确性与社会价值对齐。尽管近期取得进展，LLM 仍会延续或放大训练数据中嵌入的偏见，对公平性构成挑战。自我去偏鼓励 LLM 识别并纠正自身偏见，但依赖单个模型的内在知识可能不足以消除根深蒂固的刻板印象。为突破这一局限，我们提出 Collective Bias Mitigation（CBM），通过学习细粒度Large language models (LLMs) are increasingly deployed in public health, finance, and governance, requiring both accuracy and societal value alignment. Despite recent advances, LLMs often perpetuate or amplify bias embedded in their training data, posing challenges to fairness. While self-debiasing encourages an LLM to identify and correct its own biases, relying on a single model's intrinsic knowledge may be insufficient to address deeply ingrained stereotypes. To address this limitation, we introduce Collective Bias Mitigation (CBM), a framework that alleviates bias by learning fine-grained

arXiv ↗ 中文全文