Research map / Video / Video-language understanding
Object-centric Video Question Answering with Visual Grounding and Referring
Haochen Wang, Qirui Chen, Cilin Yan, Jiayin Cai, Xiaolong Jiang, Yao Hu, Weidi Xie, Stratis Gavves
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 STOM (track and overlay visual prompts on all frames) (ablation) | RGA3-7B · VideoInfer test | GPT-4o accuracy / score: 43.6 / 2.46 -> 46.7 / 2.60 | Improved |
| add STOM (ablation) | RGA3-7B · VideoRefer-Bench-Q | accuracy (%): 72.2 -> 74.0 | Improved |
| propagate prompts with ground-truth masks (oracle) instead of tracker (ablation) | RGA3-7B · VideoInfer test | GPT-4o accuracy: 46.7 -> 44.8 | Hurt |
| replace SAM decoder with SAM2 as grounding module (ablation) | RGA3-7B · ReVOS | J&F: 55.4 -> 58.0 | Improved |
| replace SAM decoder with SAM2 (ablation) | RGA3-7B · ReasonSeg val | gIoU: 68.5 -> 68.7 | No clear effect |
| co-train segmentation and QA data instead of segmentation only (ablation) | RGA3-3B · ReasonVOS | J&F: 48.7 -> 51.7 | Improved |
| use alpha-blended visual prompt overlay instead of region features via mask pooling | RGA3 · VideoRefer-Bench-Q | average accuracy (%): 71.9 -> 74.0 | Improved |
| scale RGA3 from 3B to 7B | RGA3 · ReVOS overall | J&F: 56.1 -> 58.0 | Improved |
| multi-task joint training of all tasks | RGA3 · individual tasks | task performance | Hurt |
| increase video duration in LongVideoBench evaluation | RGA3-7B · LongVideoBench val | accuracy (%): 72.5 -> 46.3 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- VISA-7B — VISA: Reasoning Video Object Segmentation via Large Language Model (ECCV 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Language-aware Visual Semantic Distillation for Video Question Answering (CVPR 2024 · similarity 0.91)
- Not Another Text Benchmark: Putting the “Visual" Back in Visual Question Answering for Large Video Models (NeurIPS 2026 · similarity 0.90)
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.90)
- VideoEspresso: A Large-Scale Chain-of-Thought Dataset for Fine-Grained Video Reasoning via Core Frame Selection (CVPR 2025 · similarity 0.90)
- SAMA: Towards Multi-Turn Referential Grounded Video Chat with Large Language Models (NeurIPS 2025 · similarity 0.90)
- ViCaS: A Dataset for Combining Holistic and Pixel-level Video Understanding using Captions with Grounded Segmentation (CVPR 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.
