Tracing the Evidence: Faithful Token Attribution Through Vision-Language Reasoning
cs.CV
Submitted: 2026-09-29
Updated: 2026-09-29
Project page: https://vtrace-attribution.github.io
Terminology
Sources
- AttnLRP: Attention-Aware Layer-Wise Relevance Propagation for Transformers
- Qwen3-VL Technical Report
- Advancing Multimodal Reasoning: From Optimized Cold Start to Staged Reinforcement Learning
- How Does Reasoning Flow? Tracing Attention-Induced Information Flow for Targeted RL in LLMs
- MMR1: Enhancing Multimodal Reasoning with Variance-Aware Sampling and Open Resources
- Towards Long-Horizon Interpretability: Efficient and Faithful Multi-Token Attribution for Reasoning LLMs
- RISE: Randomized Input Sampling for Explanation of Black-box Models
- GLIMPSE: Holistic Cross-Modal Explainability for Large Vision-Language Models
- HybridFlow: A Flexible and Efficient RLHF Framework
- VisualPuzzles: Decoupling Multimodal Reasoning Evaluation from Domain Knowledge
- InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency
- ReAGent: A Model-agnostic Feature Attribution Method for Generative Language Models
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models