@inproceedings{ma-etal-2026-unlocking,
title = "Unlocking Latent Medical Reasoning in {LLM}s via Inference-Time Representation and Prefix Interventions",
author = "Ma, Ming and
Li, Jiayun and
Zhao, Sendong and
Du, Yanrui and
Fan, Yuzheng and
Zhao, Danyang and
Qin, Bing and
Liu, Ting",
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.727/",
pages = "15276--15297",
ISBN = "979-8-89176-462-0",
abstract = "Large language models (LLMs) have acquired substantial medical knowledge and reasoning skills during pretraining, yet eliciting these capabilities typically requires scarce rationales or costly post-training. To address this gap, we introduce **RIPT**, a lightweight framework for data-efficient endogenous capability activation. It explicitly decomposes this objective into two complementary components: reasoning enhancement and medical knowledge elicitation. For the former, we extract steering vectors from hidden activations on a small set of high-quality paired reasoning and vanilla responses to shape LLMs' reasoning behavior. For the latter, we obtain prefix vectors via prefix tuning on simple medical QA pairs to elicit domain-specific knowledge. At inference, RIPT freezes the backbone LLM and combines steering-vector injection with prefix-vector conditioning. Rather than injecting external knowledge, RIPT aims to unlock medical reasoning capabilities already acquired by LLMs during pretraining. Experiments under limited-resource settings show that RIPT consistently improves diverse backbone LLMs and benchmarks, suggesting a generalizable pathway for lightweight medical reasoning enhancement."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ma-etal-2026-unlocking">
<titleInfo>
<title>Unlocking Latent Medical Reasoning in LLMs via Inference-Time Representation and Prefix Interventions</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ming</namePart>
<namePart type="family">Ma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiayun</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sendong</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yanrui</namePart>
<namePart type="family">Du</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuzheng</namePart>
<namePart type="family">Fan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Danyang</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bing</namePart>
<namePart type="family">Qin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ting</namePart>
<namePart type="family">Liu</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>Large language models (LLMs) have acquired substantial medical knowledge and reasoning skills during pretraining, yet eliciting these capabilities typically requires scarce rationales or costly post-training. To address this gap, we introduce **RIPT**, a lightweight framework for data-efficient endogenous capability activation. It explicitly decomposes this objective into two complementary components: reasoning enhancement and medical knowledge elicitation. For the former, we extract steering vectors from hidden activations on a small set of high-quality paired reasoning and vanilla responses to shape LLMs’ reasoning behavior. For the latter, we obtain prefix vectors via prefix tuning on simple medical QA pairs to elicit domain-specific knowledge. At inference, RIPT freezes the backbone LLM and combines steering-vector injection with prefix-vector conditioning. Rather than injecting external knowledge, RIPT aims to unlock medical reasoning capabilities already acquired by LLMs during pretraining. Experiments under limited-resource settings show that RIPT consistently improves diverse backbone LLMs and benchmarks, suggesting a generalizable pathway for lightweight medical reasoning enhancement.</abstract>
<identifier type="citekey">ma-etal-2026-unlocking</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.727/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>15276</start>
<end>15297</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Unlocking Latent Medical Reasoning in LLMs via Inference-Time Representation and Prefix Interventions
%A Ma, Ming
%A Li, Jiayun
%A Zhao, Sendong
%A Du, Yanrui
%A Fan, Yuzheng
%A Zhao, Danyang
%A Qin, Bing
%A Liu, Ting
%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 ma-etal-2026-unlocking
%X Large language models (LLMs) have acquired substantial medical knowledge and reasoning skills during pretraining, yet eliciting these capabilities typically requires scarce rationales or costly post-training. To address this gap, we introduce **RIPT**, a lightweight framework for data-efficient endogenous capability activation. It explicitly decomposes this objective into two complementary components: reasoning enhancement and medical knowledge elicitation. For the former, we extract steering vectors from hidden activations on a small set of high-quality paired reasoning and vanilla responses to shape LLMs’ reasoning behavior. For the latter, we obtain prefix vectors via prefix tuning on simple medical QA pairs to elicit domain-specific knowledge. At inference, RIPT freezes the backbone LLM and combines steering-vector injection with prefix-vector conditioning. Rather than injecting external knowledge, RIPT aims to unlock medical reasoning capabilities already acquired by LLMs during pretraining. Experiments under limited-resource settings show that RIPT consistently improves diverse backbone LLMs and benchmarks, suggesting a generalizable pathway for lightweight medical reasoning enhancement.
%U https://aclanthology.org/2026.findings-emnlp.727/
%P 15276-15297
Markdown (Informal)
[Unlocking Latent Medical Reasoning in LLMs via Inference-Time Representation and Prefix Interventions](https://aclanthology.org/2026.findings-emnlp.727/) (Ma et al., Findings 2026)
ACL
- Ming Ma, Jiayun Li, Sendong Zhao, Yanrui Du, Yuzheng Fan, Danyang Zhao, Bing Qin, and Ting Liu. 2026. Unlocking Latent Medical Reasoning in LLMs via Inference-Time Representation and Prefix Interventions. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 15276–15297, Budapest, Hungary. Association for Computational Linguistics.