@inproceedings{luo-etal-2026-surgmind,
title = "{S}urg{M}ind-{R}1: Overcoming Reasoning Collapse in Surgical Vision-Language Models via Reinforcement Learning",
author = "Luo, Yifan and
Wang, Guoming and
Xiao, Ling and
Li, Juncheng and
Tang, Siliang",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.739/",
pages = "17127--17143",
ISBN = "979-8-89176-459-0",
abstract = "Vision-Language Models (VLMs) show immense promise in medicine, yet surgical scene understanding remains formidable due to complex visual obstructions and rigorous multi-step clinical logic. Standard Supervised Fine-Tuning (SFT) imparts domain knowledge, but we identify a critical flaw{---}the ``Reasoning Collapse'' phenomenon: during answer-only SFT, models progressively lose explicit step-by-step reasoning capabilities, regressing into ``black boxes'' that generate uninterpretable shortcut answers{---}a severe risk for clinical traceability. Constructing high-quality Chain-of-Thought (CoT) datasets to mitigate this is prohibitively expensive in the surgical domain. We propose SurgMind-R1, a surgical VLM with a three-stage training pipeline. We construct Cholec-Think, a reasoning-enhanced dataset combining expert and VLM-assisted CoT annotations; apply domain-adaptive SFT and reasoning-enhanced SFT on Cholec-Think to internalize surgical logic; and introduce Reinforcement Learning via Group Relative Policy Optimization (GRPO) with rule-based rewards to enforce structured thinking paths and eliminate reasoning collapse. On SurgLite-Bench, our comprehensive benchmark, SurgMind-R1 significantly outperforms state-of-the-art VLMs on complex inference tasks. It internalizes structured thinking, maintaining high accuracy in a direct no-think mode that resolves the high-latency bottleneck of reasoning models for real-time clinical applications, and exhibits strong zero-shot out-of-distribution (OOD) generalization across unseen surgical scenarios and novel tasks, paving the way for transparent, trustworthy computer-assisted interventions. The code, benchmark, and models will be available at https://github.com/YifanLuo0103/SurgMind-R1."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="luo-etal-2026-surgmind">
<titleInfo>
<title>SurgMind-R1: Overcoming Reasoning Collapse in Surgical Vision-Language Models via Reinforcement Learning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yifan</namePart>
<namePart type="family">Luo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guoming</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ling</namePart>
<namePart type="family">Xiao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Juncheng</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Siliang</namePart>
<namePart type="family">Tang</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>Vision-Language Models (VLMs) show immense promise in medicine, yet surgical scene understanding remains formidable due to complex visual obstructions and rigorous multi-step clinical logic. Standard Supervised Fine-Tuning (SFT) imparts domain knowledge, but we identify a critical flaw—the “Reasoning Collapse” phenomenon: during answer-only SFT, models progressively lose explicit step-by-step reasoning capabilities, regressing into “black boxes” that generate uninterpretable shortcut answers—a severe risk for clinical traceability. Constructing high-quality Chain-of-Thought (CoT) datasets to mitigate this is prohibitively expensive in the surgical domain. We propose SurgMind-R1, a surgical VLM with a three-stage training pipeline. We construct Cholec-Think, a reasoning-enhanced dataset combining expert and VLM-assisted CoT annotations; apply domain-adaptive SFT and reasoning-enhanced SFT on Cholec-Think to internalize surgical logic; and introduce Reinforcement Learning via Group Relative Policy Optimization (GRPO) with rule-based rewards to enforce structured thinking paths and eliminate reasoning collapse. On SurgLite-Bench, our comprehensive benchmark, SurgMind-R1 significantly outperforms state-of-the-art VLMs on complex inference tasks. It internalizes structured thinking, maintaining high accuracy in a direct no-think mode that resolves the high-latency bottleneck of reasoning models for real-time clinical applications, and exhibits strong zero-shot out-of-distribution (OOD) generalization across unseen surgical scenarios and novel tasks, paving the way for transparent, trustworthy computer-assisted interventions. The code, benchmark, and models will be available at https://github.com/YifanLuo0103/SurgMind-R1.</abstract>
<identifier type="citekey">luo-etal-2026-surgmind</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.739/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>17127</start>
<end>17143</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T SurgMind-R1: Overcoming Reasoning Collapse in Surgical Vision-Language Models via Reinforcement Learning
%A Luo, Yifan
%A Wang, Guoming
%A Xiao, Ling
%A Li, Juncheng
%A Tang, Siliang
%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 luo-etal-2026-surgmind
%X Vision-Language Models (VLMs) show immense promise in medicine, yet surgical scene understanding remains formidable due to complex visual obstructions and rigorous multi-step clinical logic. Standard Supervised Fine-Tuning (SFT) imparts domain knowledge, but we identify a critical flaw—the “Reasoning Collapse” phenomenon: during answer-only SFT, models progressively lose explicit step-by-step reasoning capabilities, regressing into “black boxes” that generate uninterpretable shortcut answers—a severe risk for clinical traceability. Constructing high-quality Chain-of-Thought (CoT) datasets to mitigate this is prohibitively expensive in the surgical domain. We propose SurgMind-R1, a surgical VLM with a three-stage training pipeline. We construct Cholec-Think, a reasoning-enhanced dataset combining expert and VLM-assisted CoT annotations; apply domain-adaptive SFT and reasoning-enhanced SFT on Cholec-Think to internalize surgical logic; and introduce Reinforcement Learning via Group Relative Policy Optimization (GRPO) with rule-based rewards to enforce structured thinking paths and eliminate reasoning collapse. On SurgLite-Bench, our comprehensive benchmark, SurgMind-R1 significantly outperforms state-of-the-art VLMs on complex inference tasks. It internalizes structured thinking, maintaining high accuracy in a direct no-think mode that resolves the high-latency bottleneck of reasoning models for real-time clinical applications, and exhibits strong zero-shot out-of-distribution (OOD) generalization across unseen surgical scenarios and novel tasks, paving the way for transparent, trustworthy computer-assisted interventions. The code, benchmark, and models will be available at https://github.com/YifanLuo0103/SurgMind-R1.
%U https://aclanthology.org/2026.emnlp-main.739/
%P 17127-17143
Markdown (Informal)
[SurgMind-R1: Overcoming Reasoning Collapse in Surgical Vision-Language Models via Reinforcement Learning](https://aclanthology.org/2026.emnlp-main.739/) (Luo et al., EMNLP 2026)
ACL