Research map / Video / Video-language understanding
EchoBat: Echo-Vision Enhancement and Echo-Layered Sampling for Video LLMs Hallucination Mitigation
Shuai Liu, Da Chen, Yiheng Pan, Chenwei Tian, Qian Li, Chenhao Lin
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Echo-Vision Enhancement (audio-aided preference pairs, DPO) (ablation) | LLaVA-Video-7B · VideoHallucer (Ovr) | accuracy: 45.9 -> 46.3 | Improved |
| add Echo-Layered Sampling on top of EVE (ablation) | LLaVA-Video-7B · VideoHallucer (Ovr) | accuracy: 46.3 -> 46.9 | Improved |
| add Video Reverse Non-preference on top of EVE+ELS (ablation) | LLaVA-Video-7B · VideoHallucer (Ovr) | accuracy: 46.9 -> 47.2 | Improved |
| add Echo-Layered Sampling on top of EVE (ablation) | LLaVA-Video-7B · CMM video VLCorr | perception accuracy: 89.0 -> 90.5 | Improved |
| add Video Reverse Non-preference (ablation) | LLaVA-Video-7B · CMM video VLCorr | perception accuracy: 90.5 -> 92.0 | Improved |
| apply full EchoBat DPO (ablation) | LLaVA-Video-7B · CMM video LangDominance | hallucination resistance: 74.0 -> 76.0 | Improved |
| apply EchoBat | LLaVA-Video-7B · MSVD-QA | accuracy: 82.1 -> 83.7 | Improved |
| apply EchoBat | LLaVA-Video-7B · MSR-VTT QA | accuracy: 69.7 -> 71.2 | Improved |
| apply EchoBat | LLaVA-Onevision-7B · VideoHallucer (Bas) | accuracy: 60.9 -> 62.3 | Improved |
| apply EchoBat | LLaVA-Video-7B · VideoChatGPT-Bench Temporal | score: 2.2 -> 2.8 | Improved |
| use tau 0.6 instead of 0.4 for audio filtering threshold (ablation) | EchoBat · EventHallusion | accuracy: -4.1% | Hurt |
Closest papers by meaning
- Mitigating Hallucination in VideoLLMs via Temporal-Aware Activation Engineering (NeurIPS 2025 · similarity 0.92)
- Learning to Decode Against Compositional Hallucination in Video Multimodal Large Language Models (ICML 2026 · similarity 0.91)
- VidHalluc: Evaluating Temporal Hallucinations in Multimodal Large Language Models for Video Understanding (CVPR 2025 · similarity 0.91)
- MASH-VLM: Mitigating Action-Scene Hallucination in Video-LLMs through Disentangled Spatial-Temporal Representations (CVPR 2025 · similarity 0.90)
- MHBench: Demystifying Motion Hallucination in VideoLLMs (AAAI 2025 · similarity 0.90)
- Mitigating Hallucination Through Theory-Consistent Symmetric Multimodal Preference Optimization (NeurIPS 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
