analiza aiqbench
This research reframes Retrieval-Augmented Generation reliability as an observability problem by introducing Context-Driven Decomposition to distinguish between parametric priors and retrieved evidence during knowledge conflicts. The study demonstrates that standard RAG systems suffer significant accuracy drops under adversarial misinformation, while explicit decomposition techniques offer measurable robustness improvements across major models like Gemini and Claude. By isolating conflicting premises at inference time, this work provides a critical diagnostic framework for building more trustworthy AI agents in high-stakes environments where retrieval integrity is paramount.
Does RAG Know When Retrieval Is Wrong? Diagnosing Context Compliance under Knowledge Conflict
arXiv:2605.14473v4 Announce Type: replace-cross Abstract: Retrieval-Augmented Generation (RAG) is usually evaluated by whether the final answer is correct. Under knowledge conflict, this hides a key question: did the model follow retrieved evidence, rely on its parametric prior, or produce a post-hoc rationale? We study this as context compliance, the regime in which retrieved context controls the answer even when it conflicts with the model's prior knowledge. We introduce Context-Driven Decomposition (CDD), an inference-time diagnostic intervention that elicits contextual and prior answers, isolates the conflicting premise, and records a resolution trace that can be perturbed. Across Epi-Scale stress tests, TruthfulQA misconception injection, and cross-model reruns, CDD makes three behaviors visible. First, misleading retrieval can severely degrade accuracy: under a worst-case TruthfulQA misconception-injection probe, Standard RAG reaches only 15.0%. Second, better answers need not share the same mechanism: CDD improves adversarial accuracy on Gemini-2.5-Flash and shows directional gains across Claude variants, yet trace-perturbation sensitivity is high only on Gemini. Third, explicit decomposition improves controlled-conflict robustness over a conflict-aware instruction baseline on localized factual conflicts, with the clearest margins on Entity Swap (88.0% vs 79.3%) and Logical Contradiction (83.2% vs 75.4%). We frame RAG conflict handling as an observability problem.
Abstract:Retrieval-Augmented Generation (RAG) is usually evaluated by whether the final answer is correct. Under knowledge conflict, this hides a key question: did the model follow retrieved evidence, rely on its parametric prior, or produce a post-hoc rationale? We study this as context compliance, the regime in which retrieved context controls the answer even when it conflicts with the model's prior knowledge. We introduce Context-Driven Decomposition (CDD), an inference-time diagnostic intervention that elicits contextual and prior answers, isolates the conflicting premise, and records a resolution trace that can be perturbed. Across Epi-Scale stress tests, TruthfulQA misconception injection, and cross-model reruns, CDD makes three behaviors visible. First, misleading retrieval can severely degrade accuracy: under a worst-case TruthfulQA misconception-injection probe, Standard RAG reaches only 15.0%. Second, better answers need not share the same mechanism: CDD improves adversarial accuracy on Gemini-2.5-Flash and shows directional gains across Claude variants, yet trace-perturbation sensitivity is high only on Gemini. Third, explicit decomposition improves controlled-conflict robustness over a conflict-aware instruction baseline on localized factual conflicts, with the clearest margins on Entity Swap (88.0% vs 79.3%) and Logical Contradiction (83.2% vs 75.4%). We frame RAG conflict handling as an observability problem.
Preprint. 3 figures, 3 tables. Diagnostic study of context compliance in RAG under knowledge conflict; closed-API evaluations (Gemini-2.5-Flash and Claude family)
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
arXiv:2605.14473 [cs.CL]
(or arXiv:2605.14473v4 [cs.CL] for this version)
Text integral sub licența deschisă a publicației · 228 cuvinte
Subiecte conexe
An Organizational Second Brain: Building an AI That Learns From Experts
Meta Engineering · 2 sept. 2026
Presentation: From Retrieval to Reasoning: Building Production-Ready Agentic AI Systems with Knowledge Graphs
InfoQ · 12 sept. 2026
Optimizing production agents with Amazon Bedrock AgentCore Observability
AWS · 31 iul. 2026
The standard for agents you can trust: Lessons from the federal front lines
Snorkel AI Blog · 5 iun. 2026