# SurgCheck: Do Vision-Language Models Really Look at Images in Surgical VQA?

arXiv: https://arxiv.org/abs/2605.01911 · HTML: https://mlsanity.com/paper/2605.01911/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Medical vision-language models, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 18 of its 20 references that are in the map.
- First posted on arXiv 3 May 2026.

## Placed in

- Multimodal LLMs / vision-language models / [Medical vision-language models](https://mlsanity.com/mllm/medical-vision-language-models/)

## Builds on (most cited first, 18 of 18)

- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment](https://mlsanity.com/paper/2303.16634/) (2023)
- [Don't Just Assume; Look and Answer: Overcoming Priors for Visual Question Answering](https://mlsanity.com/paper/1712.00377/) (2017)
- [LLaVA-OneVision-1.5: Fully Open Framework for Democratized Multimodal Training](https://mlsanity.com/paper/2509.23661/) (2025)
- [What does CLIP know about a red circle? Visual prompt engineering for VLMs](https://mlsanity.com/paper/2304.06712/) (2023)
- [PaliGemma 2: A Family of Versatile VLMs for Transfer](https://mlsanity.com/paper/2412.03555/) (2024)
- [ViP-LLaVA: Making Large Multimodal Models Understand Arbitrary Visual Prompts](https://mlsanity.com/paper/2312.00784/) (2023)
- [Medical Visual Question Answering: A Survey](https://mlsanity.com/paper/2111.10056/) (2021)
- [Surgical-VQA: Visual Question Answering in Surgical Scenes using Transformer](https://mlsanity.com/paper/2206.11053/) (2022)
- [SurgicalGPT: End-to-End Language-Vision GPT for Visual Question Answering in Surgery](https://mlsanity.com/paper/2304.09974/) (2023)
- [Surgical-VQLA: Transformer with Gated Vision-Language Embedding for Visual Question Localized-Answering in Robotic Surgery](https://mlsanity.com/paper/2305.11692/) (2023)
- [Advancing Surgical VQA with Scene Graph Knowledge](https://mlsanity.com/paper/2312.10251/) (2023)
- [EndoChat: Grounded Multimodal Large Language Model for Endoscopic Surgery](https://mlsanity.com/paper/2501.11347/) (2025)
- [PitVQA: Image-grounded Text Embedding LLM for Visual Question Answering in Pituitary Surgery](https://mlsanity.com/paper/2405.13949/) (2024)
- [R-LLaVA: Improving Med-VQA Understanding through Visual Region of Interest](https://mlsanity.com/paper/2410.20327/) (2024)
- [Towards Holistic Surgical Scene Graph](https://mlsanity.com/paper/2507.15541/) (2025)

## Built on by (most cited first, 2 of 2)

- [STSG-VQA: Evidence-Grounded Temporal Question Answering from Surgical Spatio-Temporal Scene Graphs](https://mlsanity.com/paper/2609.08543/) (2026)
- [MoGround: Measuring and Mitigating Modality Distraction in Vision-Language Models](https://mlsanity.com/paper/2609.33431/) (2026)

## Abstract (arXiv)

Purpose: Vision-language models (VLMs) have shown promising performance in surgical visual question answering (VQA). However, existing surgical VQA datasets often contain linguistic shortcuts, where question phrasing implicitly constrains the answer space. It remains unclear whether reported performance reflects visual understanding or reliance on such linguistic shortcuts. Methods: We introduce SurgCheck, a diagnostic benchmark for quantifying linguistic shortcut reliance in surgical VQA. SurgCheck employs a paired-question design in which each surgical frame is associated with an original question containing entity names and a less-biased counterpart that removes these names while preserving identical visual content and ground-truth answers. The resulting performance gap provides a diagnostic signal of shortcut reliance. To ensure that the less-biased question remains well-defined even without entity names, four grounding cues are incorporated: bounding box, arrow, spatial position, and periphrasis. We evaluate both general-purpose and surgical-specific VLMs under zero-shot and fine-tuned settings on SurgCheck. To evaluate open-ended zero-shot responses, we introduce an LLM-as-a-judge evaluation protocol. Results: Using SurgCheck, we observe consistent performance degradation on less-biased questions across five VLMs, despite identical visual inputs. Text-only ablation reveals minimal performance drops for action and target prediction, indicating that action and target prediction is largely driven by linguistic shortcuts rather than visual reasoning. Conclusion: SurgCheck provides a controlled diagnostic framework that exposes failure modes masked by linguistic bias in existing surgical VQA benchmarks. Our findings demonstrate that strong benchmark performance does not necessarily imply faithful visual understanding, underscoring the need for bias-aware evaluation in surgical VQA.