@inproceedings{li-etal-2026-entropy-aware,
title = "Entropy-Aware Reshaping of Reinforcement Signals for Multi-Answer Reasoning",
author = "Li, Zhi and
Xu, Huidan and
Hu, Zhen and
Du, Yali and
Liu, Ying",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-acl.2001/",
pages = "40255--40268",
ISBN = "979-8-89176-395-1",
abstract = "Reinforcement learning with verifiable rewards (RLVR) is a standard post-training paradigm for large language models (LLMs), typically relying on group-wise reward and advantage normalization for stability. In set-valued multi-answer tasks, where multiple outputs may be simultaneously correct, this normalization can over-amplify a small number of early high-reward samples, suppressing learning signals from other valid generations and leading to overly concentrated updates. We propose Entropy-Aware Reshaping of Reinforcement Signals (EARS), a framework that reshapes how learning signals are normalized and aggregated. EARS uses token-level predictive entropy as an uncertainty cue to compute entropy-weighted reward statistics for advantage normalization, encouraging broader exploration and more balanced learning-signal allocation early in training. An adaptive decay schedule then anneals uncertainty-aware reweighting back to standard group normalization to ensure stable convergence. EARS further incorporates a correctness-gated multi-head process reward that provides auxiliary supervision on reasoning traces while remaining aligned with verifiable correctness. Experiments on MCTACO and MMLU-Multi using Qwen2.5-7B and Llama-3.1-8B-Instruct demonstrate consistent improvements in exact-set accuracy, training stability, and cross-dataset transfer performance on set-valued multi-answer reasoning."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="li-etal-2026-entropy-aware">
<titleInfo>
<title>Entropy-Aware Reshaping of Reinforcement Signals for Multi-Answer Reasoning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zhi</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Huidan</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhen</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yali</namePart>
<namePart type="family">Du</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ying</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Liakata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Viviane</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Jurgens</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-395-1</identifier>
</relatedItem>
<abstract>Reinforcement learning with verifiable rewards (RLVR) is a standard post-training paradigm for large language models (LLMs), typically relying on group-wise reward and advantage normalization for stability. In set-valued multi-answer tasks, where multiple outputs may be simultaneously correct, this normalization can over-amplify a small number of early high-reward samples, suppressing learning signals from other valid generations and leading to overly concentrated updates. We propose Entropy-Aware Reshaping of Reinforcement Signals (EARS), a framework that reshapes how learning signals are normalized and aggregated. EARS uses token-level predictive entropy as an uncertainty cue to compute entropy-weighted reward statistics for advantage normalization, encouraging broader exploration and more balanced learning-signal allocation early in training. An adaptive decay schedule then anneals uncertainty-aware reweighting back to standard group normalization to ensure stable convergence. EARS further incorporates a correctness-gated multi-head process reward that provides auxiliary supervision on reasoning traces while remaining aligned with verifiable correctness. Experiments on MCTACO and MMLU-Multi using Qwen2.5-7B and Llama-3.1-8B-Instruct demonstrate consistent improvements in exact-set accuracy, training stability, and cross-dataset transfer performance on set-valued multi-answer reasoning.</abstract>
<identifier type="citekey">li-etal-2026-entropy-aware</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.2001/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>40255</start>
<end>40268</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Entropy-Aware Reshaping of Reinforcement Signals for Multi-Answer Reasoning
%A Li, Zhi
%A Xu, Huidan
%A Hu, Zhen
%A Du, Yali
%A Liu, Ying
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Findings of the Association for Computational Linguistics: ACL 2026
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-395-1
%F li-etal-2026-entropy-aware
%X Reinforcement learning with verifiable rewards (RLVR) is a standard post-training paradigm for large language models (LLMs), typically relying on group-wise reward and advantage normalization for stability. In set-valued multi-answer tasks, where multiple outputs may be simultaneously correct, this normalization can over-amplify a small number of early high-reward samples, suppressing learning signals from other valid generations and leading to overly concentrated updates. We propose Entropy-Aware Reshaping of Reinforcement Signals (EARS), a framework that reshapes how learning signals are normalized and aggregated. EARS uses token-level predictive entropy as an uncertainty cue to compute entropy-weighted reward statistics for advantage normalization, encouraging broader exploration and more balanced learning-signal allocation early in training. An adaptive decay schedule then anneals uncertainty-aware reweighting back to standard group normalization to ensure stable convergence. EARS further incorporates a correctness-gated multi-head process reward that provides auxiliary supervision on reasoning traces while remaining aligned with verifiable correctness. Experiments on MCTACO and MMLU-Multi using Qwen2.5-7B and Llama-3.1-8B-Instruct demonstrate consistent improvements in exact-set accuracy, training stability, and cross-dataset transfer performance on set-valued multi-answer reasoning.
%U https://aclanthology.org/2026.findings-acl.2001/
%P 40255-40268
Markdown (Informal)
[Entropy-Aware Reshaping of Reinforcement Signals for Multi-Answer Reasoning](https://aclanthology.org/2026.findings-acl.2001/) (Li et al., Findings 2026)
ACL