HALMES: Knowing When to Intervene in LVLM Hallucination Mitigation
cs.CV, cs.AI
Submitted: 2026-10-26
Updated: 2026-10-26
Code: https://github.com/slp-tongji/Halmes
Terminology
Sources
- Qwen3-VL Technical Report
- Hallucination of Multimodal Large Language Models: A Survey
- Attention Sinks as Internal Signals for Hallucination Detection in Large Language Models
- INSIDE: LLMs' Internal States Retain the Power of Hallucination Detection
- HALC: Object Hallucination Reduction via Adaptive Focal-Contrast Decoding
- CapArena: Benchmarking and Analyzing Detailed Image Captioning in the LLM Era
- DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models
- Holistic Analysis of Hallucination in GPT-4V(ision): Bias and Interference Challenges
- MaskCD: Mitigating LVLM Hallucinations by Image Head Masked Contrastive Decoding
- Benchmarking and Improving Detail Image Caption
- MetaToken: Detecting Hallucination in Image Descriptions by Meta Classification
- Efficient Contrastive Decoding with Probabilistic Hallucination Detection - Mitigating Hallucinations in Large Vision Language Models -
- Beam Search Strategies for Neural Machine Translation
- MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models
- The Curious Case of Neural Text Degeneration
- Self-Introspective Decoding: Alleviating Hallucinations for Large Vision-Language Models
- Self-Aug: Query and Entropy Adaptive Decoding for Large Vision-Language Models
- Semantic Entropy Probes: Robust and Cheap Hallucination Detection in LLMs
- Contrastive Decoding: Open-ended Text Generation as Optimization
- Robust Hallucination Detection in LLMs via Adaptive Token Selection
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs