Evaluating Social Bias in RAG Systems: When External Context Helps and Reasoning Hurts

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Parihar, Shweta, Cheng, Lu
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Computation and Language Artificial Intelligence
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Context-Aware Counterfactual Data Augmentation for Gender Bias Mitigation in Language Models
di: Parihar, Shweta, et al.
Pubblicazione: (2026)

When RAG Hurts: Diagnosing and Mitigating Attention Distraction in Retrieval-Augmented LVLMs
di: Zhao, Beidi, et al.
Pubblicazione: (2026)

Quantization Hurts Reasoning? An Empirical Study on Quantized Reasoning Models
di: Liu, Ruikang, et al.
Pubblicazione: (2025)

When Reasoning Hurts: Source-Aware Evaluation of Frontier LLMs for Clinical SOAP Note Generation
di: Faisal, Faizan
Pubblicazione: (2026)

How Much Can RAG Help the Reasoning of LLM?
di: Liu, Jingyu, et al.
Pubblicazione: (2024)

Does Reasoning Introduce Bias? A Study of Social Bias Evaluation and Mitigation in LLM Reasoning
di: Wu, Xuyang, et al.
Pubblicazione: (2025)

Context Length Alone Hurts LLM Performance Despite Perfect Retrieval
di: Du, Yufeng, et al.
Pubblicazione: (2025)

When "Better" Prompts Hurt: Evaluation-Driven Iteration for LLM Applications
di: Commey, Daniel
Pubblicazione: (2026)

When Long Helps Short: How Context Length in Supervised Fine-tuning Affects Behavior of Large Language Models
di: Zheng, Yingming, et al.
Pubblicazione: (2025)

Towards Agentic RAG with Deep Reasoning: A Survey of RAG-Reasoning Systems in LLMs
di: Li, Yangning, et al.
Pubblicazione: (2025)

Contradiction Detection in RAG Systems: Evaluating LLMs as Context Validators for Improved Information Consistency
di: Gokul, Vignesh, et al.
Pubblicazione: (2025)

Hit-RAG: Learning to Reason with Long Contexts via Preference Alignment
di: Liu, Junming, et al.
Pubblicazione: (2026)

When to Memorize and When to Stop: Gated Recurrent Memory for Long-Context Reasoning
di: Sheng, Leheng, et al.
Pubblicazione: (2026)

Does RAG Know When Retrieval Is Wrong? Diagnosing Context Compliance under Knowledge Conflict
di: Chen, Yihang, et al.
Pubblicazione: (2026)

When Wording Steers the Evaluation: Framing Bias in LLM judges
di: Hwang, Yerin, et al.
Pubblicazione: (2026)

DeliberationBench: When Do More Voices Hurt? A Controlled Study of Multi-LLM Deliberation Protocols
di: Kaushal, Vaarunay, et al.
Pubblicazione: (2025)

CARE-RAG - Clinical Assessment and Reasoning in RAG
di: Potluri, Deepthi, et al.
Pubblicazione: (2025)

When LLMs Benchmark Themselves: Deconstructing Self-Bias in Automated Evaluation
di: Xu, Wenda, et al.
Pubblicazione: (2025)

Prompt Optimization Is a Coin Flip: Diagnosing When It Helps in Compound AI Systems
di: Zhang, Xing, et al.
Pubblicazione: (2026)

Reasoning Can Hurt the Inductive Abilities of Large Language Models
di: Jin, Haibo, et al.
Pubblicazione: (2025)

When to Trust Context: Self-Reflective Debates for Context Reliability
di: Zhou, Zeqi, et al.
Pubblicazione: (2025)

Knowledge-Graph Based RAG System Evaluation Framework
di: Dong, Sicheng, et al.
Pubblicazione: (2025)

Dual-Metric Evaluation of Social Bias in Large Language Models: Evidence from an Underrepresented Nepali Cultural Context
di: Pandey, Ashish, et al.
Pubblicazione: (2026)

When Shared Knowledge Hurts: Spectral Over-Accumulation in Model Merging
di: Li, Yayuan, et al.
Pubblicazione: (2026)

TempPerturb-Eval: On the Joint Effects of Internal Temperature and External Perturbations in RAG Robustness
di: Zhou, Yongxin, et al.
Pubblicazione: (2025)

Enhancing RAG Efficiency with Adaptive Context Compression
di: Guo, Shuyu, et al.
Pubblicazione: (2025)

RAG Makes Guardrails Unsafe? Investigating Robustness of Guardrails under RAG-style Contexts
di: She, Yining, et al.
Pubblicazione: (2025)

When Audio and Text Disagree: Revealing Text Bias in Large Audio-Language Models
di: Wang, Cheng, et al.
Pubblicazione: (2025)

Reasoning's Razor: Reasoning Improves Accuracy but Can Hurt Recall at Critical Operating Points in Safety and Hallucination Detection
di: Chegini, Atoosa, et al.
Pubblicazione: (2025)

Investigating Thinking Behaviours of Reasoning-Based Language Models for Social Bias Mitigation
di: Luo, Guoqing, et al.
Pubblicazione: (2025)

CTRL-RAG: Contrastive Likelihood Reward Based Reinforcement Learning for Context-Faithful RAG Models
di: Tan, Zhehao, et al.
Pubblicazione: (2026)

Oolong: Evaluating Long Context Reasoning and Aggregation Capabilities
di: Bertsch, Amanda, et al.
Pubblicazione: (2025)

GraphRAG-Bench: Challenging Domain-Specific Reasoning for Evaluating Graph Retrieval-Augmented Generation
di: Xiao, Yilin, et al.
Pubblicazione: (2025)

SEMA-RAG: A Self-Evolving Multi-Agent Retrieval-Augmented Generation Framework for Medical Reasoning
di: Huang, Yongfeng, et al.
Pubblicazione: (2026)

Blinded by Generated Contexts: How Language Models Merge Generated and Retrieved Contexts When Knowledge Conflicts?
di: Tan, Hexiang, et al.
Pubblicazione: (2024)

Atomic Skills are the Prerequisite: When Reinforcement Learning Synthesizes Compositional Reasoning, and When It Only Amplifies
di: Cheng, Sitao, et al.
Pubblicazione: (2025)

Case-Aware LLM-as-a-Judge Evaluation for Enterprise-Scale RAG Systems
di: Chhabra, Mukul, et al.
Pubblicazione: (2026)

ECoRAG: Evidentiality-guided Compression for Long Context RAG
di: Jeong, Yeonseok, et al.
Pubblicazione: (2025)

Does Thinking More always Help? Mirage of Test-Time Scaling in Reasoning Models
di: Ghosal, Soumya Suvra, et al.
Pubblicazione: (2025)

When Thoughts Meet Facts: Reusable Reasoning for Long-Context LMs
di: Jeong, Soyeong, et al.
Pubblicazione: (2025)