Research map / Video / Video generation
VISTA: A Test-Time Self-Improving Video Generation Agent
Do Xuan Long, Xingchen Wan, Hootan Nakhost, Chen-Yu Lee, Tomas Pfister, Sercan Ö. Arik
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 structured PromptPlanner (ablation) | VISTA · single-scene, iteration 5 | win rate vs DP (%): 35.1 -> 45.9 | Improved |
| add PromptPlanner (ablation) | VISTA · multi-scene, iteration 5 | win rate vs DP (%): 38.8 -> 46.3 | Improved |
| add pairwise tournament selection (ablation) | VISTA · single-scene, iteration 5 | win rate vs DP (%): 33.3 -> 45.9 | Improved |
| add Normal Judge to Adversarial Judge (ablation) | VISTA · multi-scene, iteration 5 | win rate vs DP (%): 26.7 -> 46.3 | Improved |
| add Adversarial Judge to Normal Judge (ablation) | VISTA · single-scene, iteration 5 | win rate vs DP (%): 17.2 -> 45.9 | Improved |
| add DTPA deep-thinking prompting agent (ablation) | VISTA · single-scene, iteration 5 | win rate vs DP (%): 37.8 -> 45.9 | Improved |
| add DTPA (ablation) | VISTA · multi-scene, iteration 5 | win rate vs DP (%): 45.2 -> 46.3 | Improved |
| scale baseline video count (VSR to VSR++) | VSR · multi-scene, iteration 5 | win-loss vs DP (%): 16.5 -> 5.9 | Hurt |
| apply Rewrite baseline instead of direct prompting | Rewrite · multi-scene, Init | win-loss vs DP (%): -16.3 | Hurt |
| use weaker T2V model Veo 2 | VISTA · single-scene | win rate vs DP (%): 15.0 -> 23.8 | Improved |
Closest papers by meaning
- SPIRAL: Self-Evolving Action-Conditioned Video Generation via Reflective Planning Agents (NeurIPS 2026 · similarity 0.86)
- VBench: Comprehensive Benchmark Suite for Video Generative Models (CVPR 2024 · similarity 0.86)
- Neuro-Symbolic Evaluation of Text-to-Video Models using Formal Verification (CVPR 2025 · similarity 0.86)
- VSTAR: Generative Temporal Nursing for Longer Dynamic Video Synthesis (ICLR 2025 · similarity 0.86)
- Improving Video Generation with Human Feedback (NeurIPS 2025 · similarity 0.86)
- Video-Bench: Human-Aligned Video Generation Benchmark (CVPR 2025 · similarity 0.86)
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.
