No-Worse Context-Aware Decoding: Preventing Neutral Regression in Context-Conditioned Generation
Fuente:
arXiv
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914485630926848 |
|---|---|
| author | Tao, Yufei Agrawal, Ameeta |
| author_facet | Tao, Yufei Agrawal, Ameeta |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_16686 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | No-Worse Context-Aware Decoding: Preventing Neutral Regression in Context-Conditioned Generation Tao, Yufei Agrawal, Ameeta Computation and Language Artificial Intelligence 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. |
| title | No-Worse Context-Aware Decoding: Preventing Neutral Regression in Context-Conditioned Generation |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2604.16686 |