<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-lin-2026-observation">
    <titleInfo>
        <title>Observation-Aware Stabilization of Reinforcement Learning for Retrieval-Augmented Generation</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Naixin</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Tong</namePart>
        <namePart type="family">Lin</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-10</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Hung-Yi</namePart>
            <namePart type="family">Lee</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Sunipa</namePart>
            <namePart type="family">Dev</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jessy</namePart>
            <namePart type="family">Li</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Desmond</namePart>
            <namePart type="family">Elliott</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Budapest, Hungary</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-462-0</identifier>
    </relatedItem>
    <abstract>Reinforcement learning (RL) for retrieval-augmented language models enables policies to interleave generated reasoning with retrieved observations. However, in Search-R1-style PPO, training may become unstable after the model has already reached strong validation performance. This paper studies late-stage collapse as a training-stability problem in search-based RL. Four metrics are proposed to quantify validation drawdown (VDD), advantage-tail expansion (ATE), actor update stress (AUS), and observation-to-actor contamination in generalized advantage estimation (GAE), denoted as OAGC. The analysis further shows how observation tokens can influence actor updates through value targets, value-model learning, and GAE. To address this issue, Observation-aware Stabilization (OAS) combines observation residual decay, old-value anchoring, and boundary smoothing. On three benchmark datasets (NQ, TriviaQA, and HotpotQA), OAS improves Search-R1 PPO under multiple retrieval budgets. In the main Top-3 setting, our method improves the exact match (EM) measure from 31.8/49.6/18.7 to 37.6/56.8/23.9 and reduces OAGC from 0.331 to 0.088, with lower search-budget saturation and no increase in the no-search rate.</abstract>
    <identifier type="citekey">zhang-lin-2026-observation</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.1242/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>26287</start>
            <end>26307</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Observation-Aware Stabilization of Reinforcement Learning for Retrieval-Augmented Generation
%A Zhang, Naixin
%A Lin, Tong
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F zhang-lin-2026-observation
%X Reinforcement learning (RL) for retrieval-augmented language models enables policies to interleave generated reasoning with retrieved observations. However, in Search-R1-style PPO, training may become unstable after the model has already reached strong validation performance. This paper studies late-stage collapse as a training-stability problem in search-based RL. Four metrics are proposed to quantify validation drawdown (VDD), advantage-tail expansion (ATE), actor update stress (AUS), and observation-to-actor contamination in generalized advantage estimation (GAE), denoted as OAGC. The analysis further shows how observation tokens can influence actor updates through value targets, value-model learning, and GAE. To address this issue, Observation-aware Stabilization (OAS) combines observation residual decay, old-value anchoring, and boundary smoothing. On three benchmark datasets (NQ, TriviaQA, and HotpotQA), OAS improves Search-R1 PPO under multiple retrieval budgets. In the main Top-3 setting, our method improves the exact match (EM) measure from 31.8/49.6/18.7 to 37.6/56.8/23.9 and reduces OAGC from 0.331 to 0.088, with lower search-budget saturation and no increase in the no-search rate.
%U https://aclanthology.org/2026.findings-emnlp.1242/
%P 26287-26307

Markdown (Informal)

[Observation-Aware Stabilization of Reinforcement Learning for Retrieval-Augmented Generation](https://aclanthology.org/2026.findings-emnlp.1242/) (Zhang & Lin, Findings 2026)

• Observation-Aware Stabilization of Reinforcement Learning for Retrieval-Augmented Generation (Zhang & Lin, Findings 2026)

ACL

• Naixin Zhang and Tong Lin. 2026. Observation-Aware Stabilization of Reinforcement Learning for Retrieval-Augmented Generation. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 26287–26307, Budapest, Hungary. Association for Computational Linguistics.