Research map / Video / Video-language understanding
Online Reasoning Video Segmentation with Just-in-Time Digital Twins
Yiqing Shen, Bohan Liu, Chenjia Li, Lalithkumar Seenivasan, Mathias Unberath
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 query-specific specialist model selection (ablation) | just-in-time digital twin framework · Video RS benchmark (Semantic L1) | J: 0.821 -> 0.865 | Improved |
| add query-specific specialist model selection (ablation) | just-in-time digital twin framework · Video RS benchmark (Temporal L3) | J: 0.664 -> 0.690 | Improved |
| add query-specific specialist model selection (ablation) | just-in-time digital twin framework · Video RS benchmark (Spatial L1) | J: 0.753 -> 0.789 | Improved |
| add digital twin update (ablation) | just-in-time digital twin framework · Video RS benchmark (Temporal L2) | J: 0.631 -> 0.705 | Improved |
| add digital twin update (ablation) | just-in-time digital twin framework · Video RS benchmark (Spatial L3) | J: 0.675 -> 0.741 | Improved |
| add digital twin update (ablation) | just-in-time digital twin framework · Video RS benchmark (Semantic L1) | J: 0.831 -> 0.865 | Improved |
| add temporal integration (sliding window) (ablation) | just-in-time digital twin framework · Video RS benchmark (Temporal L1) | J: 0.654 -> 0.721 | Improved |
| add temporal integration (sliding window) (ablation) | just-in-time digital twin framework · Video RS benchmark (Temporal L3) | J: 0.615 -> 0.690 | Improved |
| add temporal integration (sliding window) (ablation) | just-in-time digital twin framework · Video RS benchmark (Semantic L1) | J: 0.842 -> 0.865 | Improved |
| upgrade LLM-coder from gpt4o-mini to gpt4o (ablation) | just-in-time digital twin framework · Video RS benchmark (Semantic L1) | J: 0.832 -> 0.865 | Improved |
| upgrade base LLM from gpt4o-mini to gpt4o (coder gpt4o) (ablation) | just-in-time digital twin framework · Video RS benchmark (Semantic L1) | J: 0.865 -> 0.879 | Improved |
| upgrade base LLM from gpt4o-mini to gpt4o (coder gpt4o) (ablation) | just-in-time digital twin framework · Video RS benchmark (Semantic L3) | J: 0.810 -> 0.822 | Improved |
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 — VISA: Reasoning Video Object Segmentation via Large Language Model (ECCV 2024) · better on 11 of 11 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
- Fast Reasoning Segmentation for Images and Videos (CVPR 2026 · similarity 0.92)
- LISA: Reasoning Segmentation via Large Language Model (CVPR 2024 · similarity 0.88)
- SLVR: Structured Latent Visual Reasoning via Human-like Reasoning Flows (NeurIPS 2026 · similarity 0.88)
- The Devil is in Temporal Token: High Quality Video Reasoning Segmentation (CVPR 2025 · similarity 0.88)
- SciVideoBench: Benchmarking Scientific Video Reasoning in Large Multimodal Models (ICML 2026 · similarity 0.88)
- Look, Remember and Reason: Grounded Reasoning in Videos with Language Models (ICLR 2024 · 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.
