What Tokens Truly Matter? The Logit Conflation Problem in LLM Sampling

Pinlong Zhao, Huijun Tang, Pengfei Jiao, Mengyang Li


Abstract
Sampling methods for large language models select candidate tokens based on logit statistics, implicitly assuming that high logits indicate desirable outputs. We identify the Logit Conflation Problem, where a token’s logit aggregates prompt-independent factors, including linguistic fluency and parametric associations, with prompt-relevance. However, only prompt-relevance determines instruction-following quality. We propose SEAL-Sampling (Signal Extraction for Active ReLevance) to isolate this component through attention-weighted attribution. Our framework defines prompt-relevance as the causal effect of prompt content on token logits and establishes attention patterns as an efficient proxy. Experiments on LLaMA-3 demonstrate significant improvements over top-nσ, with gains of 1.8% on AlpacaEval 2.0 and 2.2% on IFEval. Furthermore, attribution scores correlate weakly with raw logits, confirming the extraction of an orthogonal signal. The method is training-free and introduces minimal latency, adding less than 12ms overhead per token.
Anthology ID:
2026.findings-acl.1841
Volume:
Findings of the Association for Computational Linguistics: ACL 2026
Month:
July
Year:
2026
Address:
San Diego, California, United States
Editors:
Maria Liakata, Viviane P. Moreira, Jiajun Zhang, David Jurgens
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
36943–36961
Language:
URL:
https://aclanthology.org/2026.findings-acl.1841/
DOI:
Bibkey:
Cite (ACL):
Pinlong Zhao, Huijun Tang, Pengfei Jiao, and Mengyang Li. 2026. What Tokens Truly Matter? The Logit Conflation Problem in LLM Sampling. In Findings of the Association for Computational Linguistics: ACL 2026, pages 36943–36961, San Diego, California, United States. Association for Computational Linguistics.
Cite (Informal):
What Tokens Truly Matter? The Logit Conflation Problem in LLM Sampling (Zhao et al., Findings 2026)
Copy Citation:
PDF:
https://aclanthology.org/2026.findings-acl.1841.pdf
Checklist:
 2026.findings-acl.1841.checklist.pdf