CoRect: Context-Aware Logit Contrast for Hidden State Rectification to Resolve Knowledge Conflicts

2026-02-09

AI Summary

This work addresses the issue in retrieval-augmented generation (RAG) where a model’s internal parametric knowledge often overrides external retrieved evidence, leading to unfaithful outputs and hallucinations. The authors propose a label-free method that identifies deep hidden states biased by parametric priors by comparing forward-pass logits with and without retrieved context. Context-aware corrections are then applied directly at these hidden layers, circumventing the limitations of decoding-time adjustments or weight editing. Evaluated on question answering and summarization tasks, the approach significantly improves output faithfulness and effectively suppresses hallucinations, outperforming multiple strong baselines.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Search and Optimization: Metareasoning and Metaheuristics
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Retrieval-Augmented Generation (RAG) often struggles with knowledge conflicts, where model-internal parametric knowledge overrides retrieved evidence, leading to unfaithful outputs. Existing approaches are often limited, relying either on superficial decoding adjustments or weight editing that necessitates ground-truth targets. Through layer-wise analysis, we attribute this failure to a parametric suppression phenomenon: specifically, in deep layers, certain FFN layers overwrite context-sensitive representations with memorized priors. To address this, we propose CoRect (Context-Aware Logit Contrast for Hidden State Rectification). By contrasting logits from contextualized and non-contextualized forward passes, CoRect identifies layers that exhibit high parametric bias without requiring ground-truth labels. It then rectifies the hidden states to preserve evidence-grounded information. Across question answering (QA) and summarization benchmarks, CoRect consistently improves faithfulness and reduces hallucinations compared to strong baselines.

Problem

• knowledge conflicts
• Retrieval-Augmented Generation
• parametric knowledge
• faithfulness
• hallucinations

Innovation

• Retrieval-Augmented Generation
• knowledge conflict
• hidden state rectification
• context-aware logit contrast
• parametric suppression