# Revealing Training Data Exposure in Vision Language Large Models via Parameter Gradients

arXiv: https://arxiv.org/abs/2606.24774 · HTML: https://mlsanity.com/paper/2606.24774/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 32 of its 56 references that are in the map.
- First posted on arXiv 23 Jun 2026.

## Builds on (most cited first, 30 of 32)

- [Scaling Laws for Neural Language Models](https://mlsanity.com/paper/2001.08361/) (2020)
- [BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation](https://mlsanity.com/paper/2201.12086/) (2022)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models](https://mlsanity.com/paper/2304.10592/) (2023)
- [Membership Inference Attacks against Machine Learning Models](https://mlsanity.com/paper/1610.05820/) (2016)
- [Extracting Training Data from Large Language Models](https://mlsanity.com/paper/2012.07805/) (2020)
- [MMBench: Is Your Multi-modal Model an All-around Player?](https://mlsanity.com/paper/2307.06281/) (2023)
- [LLaVA-Med: Training a Large Language-and-Vision Assistant for Biomedicine in One Day](https://mlsanity.com/paper/2306.00890/) (2023)
- [Vision-Language Models for Vision Tasks: A Survey](https://mlsanity.com/paper/2304.00685/) (2023)
- [Quantifying Memorization Across Neural Language Models](https://mlsanity.com/paper/2202.07646/) (2022)
- [Network Dissection: Quantifying Interpretability of Deep Visual Representations](https://mlsanity.com/paper/1704.05796/) (2017)
- [An Empirical Study of Catastrophic Forgetting in Large Language Models During Continual Fine-tuning](https://mlsanity.com/paper/2308.08747/) (2023)
- [BiomedCLIP: a multimodal biomedical foundation model pretrained from fifteen million scientific image-text pairs](https://mlsanity.com/paper/2303.00915/) (2023)
- [VLMEvalKit: An Open-Source Toolkit for Evaluating Large Multi-Modality Models](https://mlsanity.com/paper/2407.11691/) (2024)
- [Detecting Pretraining Data from Large Language Models](https://mlsanity.com/paper/2310.16789/) (2023)
- [Systematic Evaluation of Privacy Risks of Machine Learning Models](https://mlsanity.com/paper/2003.10595/) (2020)
- [PMC-CLIP: Contrastive Language-Image Pre-training using Biomedical Documents](https://mlsanity.com/paper/2303.07240/) (2023)
- [MemGuard: Defending against Black-Box Membership Inference Attacks via Adversarial Examples](https://mlsanity.com/paper/1909.10594/) (2019)
- [Scaling Up Vision-Language Pre-training for Image Captioning](https://mlsanity.com/paper/2111.12233/) (2021)
- [MedTrinity-25M: A Large-scale Multimodal Dataset with Multigranular Annotations for Medicine](https://mlsanity.com/paper/2408.02900/) (2024)
- [Outsider Oversight: Designing a Third Party Audit Ecosystem for AI Governance](https://mlsanity.com/paper/2206.04737/) (2022)
- [Fashion-Gen: The Generative Fashion Dataset and Challenge](https://mlsanity.com/paper/1806.08317/) (2018)
- [Pretraining Data Detection for Large Language Models: A Divergence-based Calibration Method](https://mlsanity.com/paper/2409.14781/) (2024)
- [ROCOv2: Radiology Objects in COntext Version 2, an Updated Multimodal Image Dataset](https://mlsanity.com/paper/2405.10004/) (2024)
- [Canary in a Coalmine: Better Membership Inference with Ensembled Adversarial Queries](https://mlsanity.com/paper/2210.10750/) (2022)
- [M^4I: Multi-modal Models Membership Inference](https://mlsanity.com/paper/2209.06997/) (2022)
- [Membership Inference Attacks against Large Vision-Language Models](https://mlsanity.com/paper/2411.02902/) (2024)
- [Con-ReCall: Detecting Pre-training Data in LLMs via Contrastive Decoding](https://mlsanity.com/paper/2409.03363/) (2024)
- [Temporal Scaling Law for Large Language Models](https://mlsanity.com/paper/2404.17785/) (2024)
- [Audit to Forget: A Unified Method to Revoke Patients' Private Data in Intelligent Healthcare](https://mlsanity.com/paper/2302.09813/) (2023)

## Abstract (arXiv)

Vision-Language Large Models (VLLMs) trained on massive crawled corpora raise pressing copyright and data-provenance concerns. These concerns are particularly acute in healthcare, where patient medical images paired with clinical reports demand rigorous privacy safeguards. However, existing training data detection methods either fail in cross-modal scenarios or rely on superficial output signals with insufficient discriminative power. We introduce GradAudit, a gradient-based auditing framework that examines internal optimization dynamics rather than treating VLLMs as black boxes. Our approach builds on a key observation: model parameters converge to regions where gradients on training samples become stable and well-aligned, whereas gradients on non-training samples remain noisy and inconsistent. By analyzing these gradient signatures, GradAudit achieves strong separability and detects genuine image-text associations learned during training, not merely individual modality membership. Empirically, across both medical and general-domain datasets, GradAudit substantially outperforms state-of-the-art baselines in both pretraining and fine-tuning VLLMs. In a case study employing copyrighted content, we show that existing training data detection methods not only underestimate the extent of unauthorized data usage, but that this underestimation becomes more pronounced as models become more recent and more advanced.