@inproceedings{wu-etal-2026-teach,
title = "Teach a Man to Fish: Self-Adaptive Reinforced Distillation for In-Context Reasoning-based Reranker",
author = "Wu, Wenwei and
LI, Youhang and
Chen, Mingyang and
Yang, Jie and
Gu, Xiaofeng and
Pan, Duyi and
Wu, Yiwen and
Zhang, Kaichen and
Wang, Wei and
Yuan, Zixuan",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.2345/",
pages = "50403--50439",
ISBN = "979-8-89176-462-0",
abstract = "Recent advances in Large Reasoning Models (LRMs) are reshaping reasoning-intensive reranking, where step-by-step reasoning enables multi-faceted relevance judgments. To scale LRMs, existing approaches distill chain-of-thought traces from a teacher LRM into a student reranker via Supervised Fine-Tuning (SFT) followed by Reinforcement Learning (RL). However, this pipeline is misaligned with the distillation objective: teacher traces are too dense for direct imitation, while RL exploration is too sparse on hard cases. To this end, we propose SARD (Self-Adaptive Reinforced Distillation), a tri-mode RL framework that recasts teacher reasoning from a fixed supervision target into an adaptive resource the student draws on only when its own reasoning falls short. Triggered by rollout outcomes, SARD organizes training into three modes: Free Exploration, where the student reasons independently; Guided Exploration, which supplies reusable in-context guidance abstracted from teacher traces when free rollouts fail; and Teacher-Guided Repair, where the teacher diagnoses failures and revises, re-selects, or appends guidance, falling back to teacher trajectories only as a last resort. Experiments on reasoning-intensive reranking benchmarks show SARD outperforms strong reranking and SFT-then-RL baselines, with clear gains on difficult instances and under-represented reasoning patterns."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wu-etal-2026-teach">
<titleInfo>
<title>Teach a Man to Fish: Self-Adaptive Reinforced Distillation for In-Context Reasoning-based Reranker</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wenwei</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Youhang</namePart>
<namePart type="family">LI</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mingyang</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jie</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiaofeng</namePart>
<namePart type="family">Gu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Duyi</namePart>
<namePart type="family">Pan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yiwen</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kaichen</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wei</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zixuan</namePart>
<namePart type="family">Yuan</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>Recent advances in Large Reasoning Models (LRMs) are reshaping reasoning-intensive reranking, where step-by-step reasoning enables multi-faceted relevance judgments. To scale LRMs, existing approaches distill chain-of-thought traces from a teacher LRM into a student reranker via Supervised Fine-Tuning (SFT) followed by Reinforcement Learning (RL). However, this pipeline is misaligned with the distillation objective: teacher traces are too dense for direct imitation, while RL exploration is too sparse on hard cases. To this end, we propose SARD (Self-Adaptive Reinforced Distillation), a tri-mode RL framework that recasts teacher reasoning from a fixed supervision target into an adaptive resource the student draws on only when its own reasoning falls short. Triggered by rollout outcomes, SARD organizes training into three modes: Free Exploration, where the student reasons independently; Guided Exploration, which supplies reusable in-context guidance abstracted from teacher traces when free rollouts fail; and Teacher-Guided Repair, where the teacher diagnoses failures and revises, re-selects, or appends guidance, falling back to teacher trajectories only as a last resort. Experiments on reasoning-intensive reranking benchmarks show SARD outperforms strong reranking and SFT-then-RL baselines, with clear gains on difficult instances and under-represented reasoning patterns.</abstract>
<identifier type="citekey">wu-etal-2026-teach</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2345/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>50403</start>
<end>50439</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Teach a Man to Fish: Self-Adaptive Reinforced Distillation for In-Context Reasoning-based Reranker
%A Wu, Wenwei
%A LI, Youhang
%A Chen, Mingyang
%A Yang, Jie
%A Gu, Xiaofeng
%A Pan, Duyi
%A Wu, Yiwen
%A Zhang, Kaichen
%A Wang, Wei
%A Yuan, Zixuan
%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 wu-etal-2026-teach
%X Recent advances in Large Reasoning Models (LRMs) are reshaping reasoning-intensive reranking, where step-by-step reasoning enables multi-faceted relevance judgments. To scale LRMs, existing approaches distill chain-of-thought traces from a teacher LRM into a student reranker via Supervised Fine-Tuning (SFT) followed by Reinforcement Learning (RL). However, this pipeline is misaligned with the distillation objective: teacher traces are too dense for direct imitation, while RL exploration is too sparse on hard cases. To this end, we propose SARD (Self-Adaptive Reinforced Distillation), a tri-mode RL framework that recasts teacher reasoning from a fixed supervision target into an adaptive resource the student draws on only when its own reasoning falls short. Triggered by rollout outcomes, SARD organizes training into three modes: Free Exploration, where the student reasons independently; Guided Exploration, which supplies reusable in-context guidance abstracted from teacher traces when free rollouts fail; and Teacher-Guided Repair, where the teacher diagnoses failures and revises, re-selects, or appends guidance, falling back to teacher trajectories only as a last resort. Experiments on reasoning-intensive reranking benchmarks show SARD outperforms strong reranking and SFT-then-RL baselines, with clear gains on difficult instances and under-represented reasoning patterns.
%U https://aclanthology.org/2026.findings-emnlp.2345/
%P 50403-50439
Markdown (Informal)
[Teach a Man to Fish: Self-Adaptive Reinforced Distillation for In-Context Reasoning-based Reranker](https://aclanthology.org/2026.findings-emnlp.2345/) (Wu et al., Findings 2026)
ACL
- Wenwei Wu, Youhang LI, Mingyang Chen, Jie Yang, Xiaofeng Gu, Duyi Pan, Yiwen Wu, Kaichen Zhang, Wei Wang, and Zixuan Yuan. 2026. Teach a Man to Fish: Self-Adaptive Reinforced Distillation for In-Context Reasoning-based Reranker. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 50403–50439, Budapest, Hungary. Association for Computational Linguistics.