<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="gupta-etal-2026-hsearch">
    <titleInfo>
        <title>HSEARCH-R1: Learning Compact Value Models for LLM-guided Hierarchical Search via RL</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Nilesh</namePart>
        <namePart type="family">Gupta</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Nasib</namePart>
        <namePart type="family">Ullah</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jinbin</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Rohit</namePart>
        <namePart type="family">Babbar</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Inderjit</namePart>
        <namePart type="given">S</namePart>
        <namePart type="family">Dhillon</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
    </relatedItem>
    <abstract>Reasoning-intensive retrieval often requires evidence whose relevance is not captured by keyword or embedding similarity alone, exposing a core limitation of traditional IR systems. An emerging direction known as LLM-guided hierarchical search addresses this limitation by letting an LLM directly traverse a semantic tree over the corpus. Such methods so far work well only with a frontier LLM in the search loop, which makes inference expensive. We propose HSearch-R1, a distillation recipe for learning compact (4B) LLMs that perform efficient LLM-guided hierarchical search. The recipe distils a frontier teacher’s per-step value estimates from its search trajectories into a much smaller LLM (Qwen3.5-4B) via RL, using a composite reward that matches the teacher’s ranking and score distribution at each step; this outcome-level RL distillation substantially outperforms supervised fine-tuning on teacher outputs. We further propose a retrieval-prior-informed variant of the search that injects a cheap retriever’s shortlist as a prior into the tree and adds robustness when the offline index is stale. The base HSearch-R1-4B matches or closely tracks frontier LLMs on the seven static BRIGHT-StackExchange subsets at roughly 50\times lower per-query cost. Adding the retrieval prior brings HSearch-R1 to 52.5 nDCG@10 on StackExchange and 45.6 across all 12 BRIGHT subsets, within 1 nDCG of the strongest published ensembled baselines while remaining substantially cheaper.</abstract>
    <identifier type="citekey">gupta-etal-2026-hsearch</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.2494/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>56623</start>
            <end>56642</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T HSEARCH-R1: Learning Compact Value Models for LLM-guided Hierarchical Search via RL
%A Gupta, Nilesh
%A Ullah, Nasib
%A Zhang, Jinbin
%A Babbar, Rohit
%A Dhillon, Inderjit S.
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F gupta-etal-2026-hsearch
%X Reasoning-intensive retrieval often requires evidence whose relevance is not captured by keyword or embedding similarity alone, exposing a core limitation of traditional IR systems. An emerging direction known as LLM-guided hierarchical search addresses this limitation by letting an LLM directly traverse a semantic tree over the corpus. Such methods so far work well only with a frontier LLM in the search loop, which makes inference expensive. We propose HSearch-R1, a distillation recipe for learning compact (4B) LLMs that perform efficient LLM-guided hierarchical search. The recipe distils a frontier teacher’s per-step value estimates from its search trajectories into a much smaller LLM (Qwen3.5-4B) via RL, using a composite reward that matches the teacher’s ranking and score distribution at each step; this outcome-level RL distillation substantially outperforms supervised fine-tuning on teacher outputs. We further propose a retrieval-prior-informed variant of the search that injects a cheap retriever’s shortlist as a prior into the tree and adds robustness when the offline index is stale. The base HSearch-R1-4B matches or closely tracks frontier LLMs on the seven static BRIGHT-StackExchange subsets at roughly 50\times lower per-query cost. Adding the retrieval prior brings HSearch-R1 to 52.5 nDCG@10 on StackExchange and 45.6 across all 12 BRIGHT subsets, within 1 nDCG of the strongest published ensembled baselines while remaining substantially cheaper.
%U https://aclanthology.org/2026.emnlp-main.2494/
%P 56623-56642

Markdown (Informal)

[HSEARCH-R1: Learning Compact Value Models for LLM-guided Hierarchical Search via RL](https://aclanthology.org/2026.emnlp-main.2494/) (Gupta et al., EMNLP 2026)

• HSEARCH-R1: Learning Compact Value Models for LLM-guided Hierarchical Search via RL (Gupta et al., EMNLP 2026)

ACL

• Nilesh Gupta, Nasib Ullah, Jinbin Zhang, Rohit Babbar, and Inderjit S Dhillon. 2026. HSEARCH-R1: Learning Compact Value Models for LLM-guided Hierarchical Search via RL. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 56623–56642, Budapest, Hungary. Association for Computational Linguistics.