@inproceedings{liang-etal-2026-rlhs,
title = "{RLHS}: Mitigating Misalignment in {RLHF} with Hindsight Simulation",
author = "Liang, Kaiqu and
Hu, Haimin and
Liu, Ryan and
Griffiths, Thomas L. and
Fisac, Jaime Fern{\'a}ndez",
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.556/",
pages = "11457--11483",
ISBN = "979-8-89176-395-1",
abstract = "While Reinforcement Learning from Human Feedback (RLHF) has shown promise in aligning generative AI, we present empirical evidence that it can also cause severe, systematic misalignment. We hypothesize that this stems from evaluator feedback depending on downstream outcome predictions (foresight) that can be influenced by the AI{'}s output, inducing Goodhart{'}s law dynamics. We present a theoretical analysis showing that conditioning evaluator feedback on downstream observations (hindsight) inhibits this effect by decoupling the alignment signal from potentially compromised predictions{---}crucially, the result holds even if the observed outcomes are sampled from the AI{'}s own world model. Building on this insight, we introduce Reinforcement Learning from Hindsight Simulation (RLHS), which presents plausible simulated outcomes to evaluators before eliciting feedback. We validate RLHS across three consultancy settings{---}marketplace interactions, restaurant recommendations, and online course advising{---}using both online (PPO) and offline (DPO) fine-tuning methods, and show that it substantially improves alignment over RLHF in experiments and human evaluations. We perform post-hoc benchmark evaluations on TruthfulQA, HaluEval, and TrustLLM, finding that even after single-task fine-tuning, RLHF misalignment persists, while RLHS consistently outperforms baselines and demonstrates strong out-of-domain generalization."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="liang-etal-2026-rlhs">
<titleInfo>
<title>RLHS: Mitigating Misalignment in RLHF with Hindsight Simulation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Kaiqu</namePart>
<namePart type="family">Liang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haimin</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ryan</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Thomas</namePart>
<namePart type="given">L</namePart>
<namePart type="family">Griffiths</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jaime</namePart>
<namePart type="given">Fernández</namePart>
<namePart type="family">Fisac</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>While Reinforcement Learning from Human Feedback (RLHF) has shown promise in aligning generative AI, we present empirical evidence that it can also cause severe, systematic misalignment. We hypothesize that this stems from evaluator feedback depending on downstream outcome predictions (foresight) that can be influenced by the AI’s output, inducing Goodhart’s law dynamics. We present a theoretical analysis showing that conditioning evaluator feedback on downstream observations (hindsight) inhibits this effect by decoupling the alignment signal from potentially compromised predictions—crucially, the result holds even if the observed outcomes are sampled from the AI’s own world model. Building on this insight, we introduce Reinforcement Learning from Hindsight Simulation (RLHS), which presents plausible simulated outcomes to evaluators before eliciting feedback. We validate RLHS across three consultancy settings—marketplace interactions, restaurant recommendations, and online course advising—using both online (PPO) and offline (DPO) fine-tuning methods, and show that it substantially improves alignment over RLHF in experiments and human evaluations. We perform post-hoc benchmark evaluations on TruthfulQA, HaluEval, and TrustLLM, finding that even after single-task fine-tuning, RLHF misalignment persists, while RLHS consistently outperforms baselines and demonstrates strong out-of-domain generalization.</abstract>
<identifier type="citekey">liang-etal-2026-rlhs</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.556/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>11457</start>
<end>11483</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T RLHS: Mitigating Misalignment in RLHF with Hindsight Simulation
%A Liang, Kaiqu
%A Hu, Haimin
%A Liu, Ryan
%A Griffiths, Thomas L.
%A Fisac, Jaime Fernández
%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 liang-etal-2026-rlhs
%X While Reinforcement Learning from Human Feedback (RLHF) has shown promise in aligning generative AI, we present empirical evidence that it can also cause severe, systematic misalignment. We hypothesize that this stems from evaluator feedback depending on downstream outcome predictions (foresight) that can be influenced by the AI’s output, inducing Goodhart’s law dynamics. We present a theoretical analysis showing that conditioning evaluator feedback on downstream observations (hindsight) inhibits this effect by decoupling the alignment signal from potentially compromised predictions—crucially, the result holds even if the observed outcomes are sampled from the AI’s own world model. Building on this insight, we introduce Reinforcement Learning from Hindsight Simulation (RLHS), which presents plausible simulated outcomes to evaluators before eliciting feedback. We validate RLHS across three consultancy settings—marketplace interactions, restaurant recommendations, and online course advising—using both online (PPO) and offline (DPO) fine-tuning methods, and show that it substantially improves alignment over RLHF in experiments and human evaluations. We perform post-hoc benchmark evaluations on TruthfulQA, HaluEval, and TrustLLM, finding that even after single-task fine-tuning, RLHF misalignment persists, while RLHS consistently outperforms baselines and demonstrates strong out-of-domain generalization.
%U https://aclanthology.org/2026.findings-acl.556/
%P 11457-11483
Markdown (Informal)
[RLHS: Mitigating Misalignment in RLHF with Hindsight Simulation](https://aclanthology.org/2026.findings-acl.556/) (Liang et al., Findings 2026)
ACL
- Kaiqu Liang, Haimin Hu, Ryan Liu, Thomas L. Griffiths, and Jaime Fernández Fisac. 2026. RLHS: Mitigating Misalignment in RLHF with Hindsight Simulation. In Findings of the Association for Computational Linguistics: ACL 2026, pages 11457–11483, San Diego, California, United States. Association for Computational Linguistics.