@inproceedings{wang-etal-2026-dont,
title = "Don{'}t Tell the Answer, Truly Guide the Reasoning During {RL} Rollouts",
author = "Wang, Xinyi and
Han, Jinyi and
Jiang, Zishang and
li, Tingyun and
Liang, Jiaqing and
Jiang, Sihang and
Dai, Zhaoqian and
Shuguang, Ma and
Yu, Fei and
Xiao, Yanghua",
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.169/",
pages = "3437--3455",
ISBN = "979-8-89176-395-1",
abstract = "Reinforcement learning (RL) has emerged as a key approach for improving long chain-of-thought (CoT) reasoning in large language models (LLMs). However, existing methods such as GRPO often break down when task difficulty exceeds the model{'}s capacity, resulting in sparse rewards and inefficient training. While prior work attempts to address this issue using off-policy data, it frequently introduces distributional mismatch, leading to unstable policy updates.In this work, we identify a fundamental issue underlying these limitations, which we term *low training affinity*, and propose **Affinity**, the first quantitative metric for measuring the compatibility between external guidance and a model{'}s intrinsic policy. Based on this insight, we introduce **HINT**, an adaptive framework designed to enhance reasoning performance while explicitly preserving high Affinity.HINT consists of two key components. First, instead of providing partial answers, it introduces **Meta-Hints**, which serve as abstract cognitive scaffolding that guides the model to independently construct solutions. Second, we propose **Affinity-Aware Policy Optimization (AAPO)**, which dynamically adjusts the learning objective based on the Affinity signal to ensure stable training.Extensive experiments across diverse benchmarks demonstrate that HINT consistently outperforms strong baselines, while achieving improved training stability and robust generalization to out-of-distribution tasks. Code is available at: https://github.com/ViviqwerAsd/HINT"
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2026-dont">
<titleInfo>
<title>Don’t Tell the Answer, Truly Guide the Reasoning During RL Rollouts</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xinyi</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jinyi</namePart>
<namePart type="family">Han</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zishang</namePart>
<namePart type="family">Jiang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tingyun</namePart>
<namePart type="family">li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiaqing</namePart>
<namePart type="family">Liang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sihang</namePart>
<namePart type="family">Jiang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhaoqian</namePart>
<namePart type="family">Dai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ma</namePart>
<namePart type="family">Shuguang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fei</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yanghua</namePart>
<namePart type="family">Xiao</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 (RL) has emerged as a key approach for improving long chain-of-thought (CoT) reasoning in large language models (LLMs). However, existing methods such as GRPO often break down when task difficulty exceeds the model’s capacity, resulting in sparse rewards and inefficient training. While prior work attempts to address this issue using off-policy data, it frequently introduces distributional mismatch, leading to unstable policy updates.In this work, we identify a fundamental issue underlying these limitations, which we term *low training affinity*, and propose **Affinity**, the first quantitative metric for measuring the compatibility between external guidance and a model’s intrinsic policy. Based on this insight, we introduce **HINT**, an adaptive framework designed to enhance reasoning performance while explicitly preserving high Affinity.HINT consists of two key components. First, instead of providing partial answers, it introduces **Meta-Hints**, which serve as abstract cognitive scaffolding that guides the model to independently construct solutions. Second, we propose **Affinity-Aware Policy Optimization (AAPO)**, which dynamically adjusts the learning objective based on the Affinity signal to ensure stable training.Extensive experiments across diverse benchmarks demonstrate that HINT consistently outperforms strong baselines, while achieving improved training stability and robust generalization to out-of-distribution tasks. Code is available at: https://github.com/ViviqwerAsd/HINT</abstract>
<identifier type="citekey">wang-etal-2026-dont</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.169/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>3437</start>
<end>3455</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Don’t Tell the Answer, Truly Guide the Reasoning During RL Rollouts
%A Wang, Xinyi
%A Han, Jinyi
%A Jiang, Zishang
%A li, Tingyun
%A Liang, Jiaqing
%A Jiang, Sihang
%A Dai, Zhaoqian
%A Shuguang, Ma
%A Yu, Fei
%A Xiao, Yanghua
%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 wang-etal-2026-dont
%X Reinforcement learning (RL) has emerged as a key approach for improving long chain-of-thought (CoT) reasoning in large language models (LLMs). However, existing methods such as GRPO often break down when task difficulty exceeds the model’s capacity, resulting in sparse rewards and inefficient training. While prior work attempts to address this issue using off-policy data, it frequently introduces distributional mismatch, leading to unstable policy updates.In this work, we identify a fundamental issue underlying these limitations, which we term *low training affinity*, and propose **Affinity**, the first quantitative metric for measuring the compatibility between external guidance and a model’s intrinsic policy. Based on this insight, we introduce **HINT**, an adaptive framework designed to enhance reasoning performance while explicitly preserving high Affinity.HINT consists of two key components. First, instead of providing partial answers, it introduces **Meta-Hints**, which serve as abstract cognitive scaffolding that guides the model to independently construct solutions. Second, we propose **Affinity-Aware Policy Optimization (AAPO)**, which dynamically adjusts the learning objective based on the Affinity signal to ensure stable training.Extensive experiments across diverse benchmarks demonstrate that HINT consistently outperforms strong baselines, while achieving improved training stability and robust generalization to out-of-distribution tasks. Code is available at: https://github.com/ViviqwerAsd/HINT
%U https://aclanthology.org/2026.findings-acl.169/
%P 3437-3455
Markdown (Informal)
[Don’t Tell the Answer, Truly Guide the Reasoning During RL Rollouts](https://aclanthology.org/2026.findings-acl.169/) (Wang et al., Findings 2026)
ACL
- Xinyi Wang, Jinyi Han, Zishang Jiang, Tingyun li, Jiaqing Liang, Sihang Jiang, Zhaoqian Dai, Ma Shuguang, Fei Yu, and Yanghua Xiao. 2026. Don’t Tell the Answer, Truly Guide the Reasoning During RL Rollouts. In Findings of the Association for Computational Linguistics: ACL 2026, pages 3437–3455, San Diego, California, United States. Association for Computational Linguistics.