Research map / Video / Video-language understanding
SPOT: Spatiotemporal Prompt Optimization for Motion-Stabilized MLLM-Guided Video Segmentation
Jiayi Fan, Zheyun Qin, Xiaoming Xi, Xiushan Nie, Yilong Yin
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 Brownian bridge loss (ablation) | SPOT-7B · Ref-YouTube-VOS | J&F: 65.2 -> 70.5 | Improved |
| add prompt quality-aware loss (ablation) | SPOT-7B · Ref-YouTube-VOS | J&F: 62.7 -> 70.5 | Improved |
| add text alignment loss (ablation) | SPOT-7B · Ref-YouTube-VOS | J&F: 67.8 -> 70.5 | Improved |
| original SAM with SPOT losses vs SAM 2 baseline (ablation) | MLLM + SAM · Ref-YouTube-VOS | J&F: 66.9 -> 70.5 | Improved |
| variance-adaptive weighting instead of constant weighting (ablation) | Brownian bridge loss · Ref-YouTube-VOS | J&F: 68.7 -> 70.5 | Improved |
| supervise intermediate frames instead of endpoints only (ablation) | Brownian bridge loss · Ref-YouTube-VOS | J&F: 67.3 -> 70.5 | Improved |
| add geometric (BCE) point supervision (ablation) | prompt quality loss · Ref-YouTube-VOS | J&F: 67.8 -> 70.5 | Improved |
| add bounding box loss (ablation) | prompt quality loss · Ref-YouTube-VOS | J&F: 68.3 -> 70.5 | Improved |
| scale SPOT from 7B to 13B | SPOT · ReVOS overall | J&F: 52.7 -> 54.8 | Improved |
| SPOT vs prior best on Ref-YouTube-VOS | SPOT-7B · Ref-YouTube-VOS | J&F: 69.2 -> 70.5 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- VISA-13B — VISA: Reasoning Video Object Segmentation via Large Language Model (ECCV 2024) · better on 5 of 5 shared results
- DTOS-9B — DTOS: Dynamic Time Object Sensing with Large Multimodal Model (CVPR 2025) · better on 3 of 3 shared results
- SAMWISE — SAMWISE: Infusing Wisdom in SAM2 for Text-Driven Video Segmentation (CVPR 2025) · better on 3 of 3 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
- MotionBench: Benchmarking and Improving Fine-grained Video Motion Understanding for Vision Language Models (CVPR 2025 · similarity 0.89)
- The Devil is in Temporal Token: High Quality Video Reasoning Segmentation (CVPR 2025 · similarity 0.88)
- Exploiting Temporal State Space Sharing for Video Semantic Segmentation (CVPR 2025 · similarity 0.87)
- Unleashing the Temporal-Spatial Reasoning Capacity of GPT for Training-Free Audio and Language Referenced Video Object Segmentation (AAAI 2025 · similarity 0.87)
- MVBench: A Comprehensive Multi-modal Video Understanding Benchmark (CVPR 2024 · similarity 0.87)
- PixFoundation 2.0: Do Video Multi-Modal LLMs Use Motion in Visual Grounding? (NeurIPS 2026 · similarity 0.87)
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.
