Model Releases
Revealing Multi-View Hallucination in Large Vision-Language Models
arXiv:2603.23934v2 Announce Type: replace-cross Abstract: Large vision-language models (LVLMs) are increasingly being applied to multi-view image inputs captured from diverse viewpoints. Despite this
arXiv:2603.23934v2 Announce Type: replace-cross Abstract: Large vision-language models (LVLMs) are increasingly being applied to multi-view image inputs captured from diverse viewpoints. Despite this growing use, current LVLMs often generate incorrect responses due to visual interference from non-target instances or viewpoints, a phenomenon we term multi-view hallucination (MVH). To systematically analyze this problem, we construct MVH-Bench, a benchmark comprising 4.8k question-answer pairs targeting two types of hallucination: cross-instance and cross-view. Empirical results show that MVH is prevalent across recent LVLMs. To address this issue, we propose Reference Shift Contrastive Decoding (RSCD), a training-free decoding technique that suppresses visual interference by generating negative logits through attention masking. Experiments on MVH-Bench with LLaVA-OneVision and Qwen2.5-VL demonstrate that RSCD improves performance by up to 25.7 and 94.8 points over existing hallucination mitigation methods, highlighting the effectiveness of RSCD.
Related
- Benchmarking Deflection and Hallucination in Large Vision-Language Models
- Fine-Grained Multi Image Object Hallucination Benchmark
- ORCA: An Agentic Reasoning Framework for Hallucination and Adversarial Robustness in Vision-Language Models
- Mitigating Hallucination in Vision-Language Models through Barrier-Regulated Adaptive Closed-form Steering
Source: arXiv cs.AI | 2026-09-02