# GAZE: Grounded Agentic Zero-shot Evaluation with Viewer-Level Tools and Literature Retrieval on Rare Brain MRI

arXiv: https://arxiv.org/abs/2605.00876 · HTML: https://mlsanity.com/paper/2605.00876/

## Key facts

- Cited by 1 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 16 of its 26 references that are in the map.
- First posted on arXiv 25 Apr 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, 16 of 16)

- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [Dense Passage Retrieval for Open-Domain Question Answering](https://mlsanity.com/paper/2004.04906/) (2020)
- [Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrieval](https://mlsanity.com/paper/2007.00808/) (2020)
- [Capabilities of Gemini Models in Medicine](https://mlsanity.com/paper/2404.18416/) (2024)
- [Advancing Multimodal Medical Capabilities of Gemini](https://mlsanity.com/paper/2405.03162/) (2024)
- [MedRAX: Medical Reasoning Agent for Chest X-ray](https://mlsanity.com/paper/2502.02673/) (2025)
- [ChestX-Reasoner: Advancing Radiology Foundation Models with Reasoning through Step-by-Step Verification](https://mlsanity.com/paper/2504.20930/) (2025)
- [CT-Agent: A Multimodal-LLM Agent for 3D CT Radiology Question Answering](https://mlsanity.com/paper/2505.16229/) (2025)
- [NOVA: A Benchmark for Anomaly Localization and Clinical Reasoning in Brain MRI](https://mlsanity.com/paper/2505.14064/) (2025)
- [CXR-Agent: Vision-language models for chest X-ray interpretation with uncertainty aware radiology reporting](https://mlsanity.com/paper/2407.08811/) (2024)
- [VoxelPrompt: A Vision Agent for End-to-End Medical Image Analysis](https://mlsanity.com/paper/2410.08397/) (2024)
- [AURA: A Multi-Modal Medical Agent for Understanding, Reasoning & Annotation](https://mlsanity.com/paper/2507.16940/) (2025)
- [Agentic Systems in Radiology: Design, Applications, Evaluation, and Challenges](https://mlsanity.com/paper/2510.09404/) (2025)
- [MedOpenClaw and MedFlowBench: Auditing Medical Agents in Full-Study Workflows](https://mlsanity.com/paper/2603.24649/) (2026)
- [Learning to reason about rare diseases through retrieval-augmented agents](https://mlsanity.com/paper/2511.04720/) (2025)
- [Radiologist Copilot: An Agentic Framework Orchestrating Specialized Tools for Reliable Radiology Reporting](https://mlsanity.com/paper/2512.02814/) (2025)

## Built on by (most cited first, 1 of 1)

- [Volumetric Radiology AI in the Era of Multimodal Large Language Models](https://mlsanity.com/paper/2608.20549/) (2026)

## Abstract (arXiv)

Vision-language models (VLMs) read an image and produce text in a single forward pass, whereas radiologists typically inspect an image several times and consult the literature before writing a report. We introduce GAZE (Grounded Agentic Zero-shot Evaluation), a framework that lets a medical VLM work in this iterative way by calling viewer-level tools (zoom, windowing, contrast, edge detection) and two retrieval tools backed by the U.S. National Library of Medicine (PubMed for medical literature, Open-i for radiological images), with structured outputs validated against a schema and full tool-call traces recorded for auditability. On NOVA, a benchmark of 906 brain MRI cases covering 281 rare neurological conditions, GAZE reaches 58.2 mean average precision (mAP) at intersection-over-union (IoU) 0.3 for lesion localisation and 34.9% Top-1 diagnostic accuracy under a joint protocol that scores captioning, diagnosis, and localisation from the image alone, without task-specific fine-tuning. Before any tool is used, structured prompting and schema-validated outputs already improve over the published Gemini 2.0 Flash baseline (20.2 to 29.4 mAP@0.3), so framework design is itself an experimental variable. Tool use helps rare pathologies disproportionately: the fraction of cases with IoU > 0.3 rises from 17% to 58% for diagnoses with three or fewer examples versus 25% to 68% for common conditions (≥10 cases), with gains tracking engagement (Gemini 3 Flash: Cohen's d = 0.79, 11.8 tool calls per case; Gemini 2.0 Flash: tools used in 8.2% of cases, no significant benefit). Retrieval ablations additionally reveal a model-dependent trade-off in which gains in diagnosis can coincide with losses in localisation, reinforcing the case for joint evaluation of diagnosis, localisation, and captioning in medical VLMs.