Research map / Video / Video-language understanding
UVLM: Benchmarking Video Language Model for Underwater World Understanding
Xizhe Xue, Yang Zhou, Dawei Yan, Lijie Tao, Junjie Li, Ying Li, Haokui Zhang, Rong Xiao
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 |
|---|---|---|---|
| fine-tune on UVLM | VideoLLaMA3-7B · UVLM test set | Overall Accuracy: 62.70 -> 73.04 | Improved |
| fine-tune on UVLM | Qvis2.5-2B · UVLM test set | Overall Accuracy: 53.06 -> 68.85 | Improved |
| fine-tune on UVLM | InternVL2.5-1B · UVLM test set | Overall Accuracy: 46.73 -> 59.14 | Improved |
| fine-tune on UVLM | Qwen2.5VL-2B · UVLM test set | Overall Accuracy: 52.97 -> 58.44 | Improved |
| fine-tune on UVLM | Qwen2.5VL-7B · UVLM test set | Overall Accuracy: 63.57 -> 68.08 | Improved |
| fine-tune on UVLM | InternVL2.5-8B · UVLM test set | Fine-grained taxonomic classification: 36.63 -> 43.94 | Improved |
| fine-tune on UVLM | VideoLLaMA3-7B · VideoMME | accuracy: 65.6 -> 65.8 | No clear effect |
| fine-tune on UVLM | InternVL2.5-1B · Perception Test | accuracy: 59.8 -> 59.6 | No clear effect |
| fine-tune on UVLM | VideoLLaMA3-2B · VideoMME | accuracy: 59.4 -> 59.2 | No clear effect |
| evaluate with Gemini judge instead of GPT-4o | VideoLLaMA3-7B + UVLM · UVLM test set | Overall: 49.59 -> 65.57 | Improved |
| add prior knowledge (bounding boxes, taxonomy) to GPT-4o annotation (ablation) | GPT-4o annotation pipeline · annotation accuracy objective labels | accuracy (%): 45 -> 91 | Improved |
| add manual correction after prior knowledge (ablation) | GPT-4o annotation pipeline · annotation accuracy objective labels | accuracy (%): 91 -> 100 | Improved |
Closest papers by meaning
- NAUTILUS: A Large Multimodal Model for Underwater Scene Understanding (NeurIPS 2025 · similarity 0.88)
- WebUOT-1M: Advancing Deep Underwater Object Tracking with A Million-Scale Benchmark (NeurIPS 2024 · similarity 0.86)
- UVEB: A Large-scale Benchmark and Baseline Towards Real-World Underwater Video Enhancement (CVPR 2024 · similarity 0.86)
- Diving into Underwater: Segment Anything Model Guided Underwater Salient Instance Segmentation and A Large-scale Dataset (ICML 2024 · similarity 0.85)
- CoralVQA: A Large-Scale Visual Question Answering Dataset for Coral Reef Image Understanding (NeurIPS 2025 · similarity 0.85)
- A Unified Image-Dense Annotation Generation Model for Underwater Scenes (CVPR 2025 · similarity 0.84)
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.
