Interpreting and Enhancing Emotional Circuits in Large Vision-Language Models via Cross-Modal Information Flow
cs.CV, cs.AI
Submitted: 2026-05-21
Updated: 2026-09-24
Terminology
Sources
- LLaVA-OneVision-1.5: Fully Open Framework for Democratized Multimodal Training
- Causal Tracing of Object Representations in Large Vision Language Models: Mechanistic Interpretability and Hallucination Mitigation
- Evaluating Object Hallucination in Large Vision-Language Models
- The Hidden Life of Tokens: Reducing Hallucination of Large Vision-Language Models via Visual Information Steering
- What Do VLMs NOTICE? A Mechanistic Interpretability Pipeline for Gaussian-Noise-free Text-Image Corruption and Evaluation
- Emotion-Qwen: A Unified Framework for Emotion and Vision Understanding
- OV-MER: Towards Open-Vocabulary Multimodal Emotion Recognition
- Interpreting and Editing Vision-Language Representations to Mitigate Hallucinations
- AffectGPT: A New Dataset, Model, and Benchmark for Emotion Understanding with Multimodal Large Language Models
- AffectGPT-R1: Leveraging Reinforcement Learning for Open-Vocabulary Multimodal Emotion Recognition
- Towards Interpreting Visual Information Processing in Vision-Language Models
- Towards Emotional Support Dialog Systems
- Reducing Hallucinations in Vision-Language Models via Latent Space Steering
- The Alignment Problem from a Deep Learning Perspective
- Same Task, Different Circuits: Disentangling Modality-Specific Mechanisms in VLMs
- Alignment and Safety in Large Language Models: Safety Mechanisms, Training Paradigms, and Emerging Challenges
- Vision-Language Models Create Cross-Modal Task Representations
- Object Hallucination in Image Captioning
- Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networks
- Proximal Policy Optimization Algorithms
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