No-Worse Context-Aware Decoding: Preventing Neutral Regression in Context-Conditioned Generation
Not on the map. Why?
Abstract (arXiv)
Large language models (LLMs) can answer questions and summarize documents when conditioned on external contexts (e.g., retrieved evidence), yet context use remains unreliable: models may overwrite an already-correct output (neutral regression) even when the context is non-informative. We formalize neutral regression as a do-no-harm requirement and quantify it by measuring accuracy drops on baseline-correct items under answer-consistent contexts. We propose No-Worse Context-Aware Decoding (NWCAD), a decode-time adapter built on a two-stream setup with a two-stage gate: it backs off to no-context decoding when the context is non-informative, and otherwise uses context-conditioned decoding with a CAD-style fallback under uncertainty. We evaluate NWCAD on benchmarks that separate do-no-harm reliability from context utilization (accuracy gains on genuinely helpful contexts). NWCAD prevents neutral regression on baseline-correct items while preserving strong context-driven accuracy on helpful contexts.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 19 of its 22 references that are in the map. First posted on arXiv 17 Apr 2026.
Builds on
19 of 22 references.
References · 19
- 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,160
- 2018HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answeringcited by 3,686
- 2023FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generationcited by 1,259
- 2022When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memoriescited by 1,103
- 2019Latent Retrieval for Weakly Supervised Open Domain Question Answeringcited by 691
- 2019Evaluating the Factual Consistency of Abstractive Text Summarizationcited by 578
- 2022Dynamic Prompt Learning via Policy Gradient for Semi-structured Mathematical Reasoningcited by 347
- 2024MiniCheck: Efficient Fact-Checking of LLMs on Grounding Documentscited by 251
- 2023ExpertQA: Expert-Curated Questions and Attributed Answerscited by 96
- 2024TofuEval: Evaluating Hallucinations of LLMs on Topic-Focused Dialogue Summarizationcited by 66
- 2024Uncertainty-Based Abstention in LLMs Improves Safety and Reduces Hallucinationscited by 51
- 2024AdaCAD: Adaptively Decoding to Balance Conflicts between Contextual and Parametric Knowledgecited by 35
7 more
- 2024Fine-Grained Natural Language Inference Based Faithfulness Evaluation for Diverse Summarisation Taskscited by 23
- 2024SFR-RAG: Towards Contextually Faithful LLMscited by 19
- 2025CoCoA: Confidence and Context-Aware Adaptive Decoding for Resolving Knowledge Conflicts in Large Language Modelscited by 9
- 2024FactTest: Factuality Testing in Large Language Models with Finite-Sample and Distribution-Free Guaranteescited by 6
- 2025"Lost-in-the-Later": Framework for Quantifying Contextual Grounding in Large Language Modelscited by 5
- 2025Replace, Don't Expand: Mitigating Context Dilution in Multi-Hop RAG via Fixed-Budget Evidence Assemblycited by 3
- 2025Dynamic Context Selection for Retrieval-Augmented Generation: Mitigating Distractors and Positional Biascited by 3
Built on by
Citing papers per year: 2026: 1.
1 paper.