Research map / Video / Video generation
Diverse Video Generation with Determinantal Point Process-Guided Policy Optimization
Tahira Kazimi, Connor Dunlop, Pinar Yanardag
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 DPP-GRPO prompt policy to Wan2.1 | Wan2.1 · VBench prompts (Wan2.1) | TCE: 19.76 -> 31.95 | Improved |
| add DPP-GRPO prompt policy to Wan2.1 | Wan2.1 · VBench prompts (Wan2.1) | TIE: 39.7 -> 49.09 | Improved |
| add DPP-GRPO prompt policy to CogVideoX | CogVideoX · VBench prompts (CogVideoX) | TCE: 22.21 -> 27.59 | Improved |
| add DPP-GRPO prompt policy to CogVideoX | CogVideoX · VBench prompts (CogVideoX) | CLIP: 0.292 -> 0.310 | Improved |
| rewrite prompts with GPT-5 instead of original prompts | Wan2.1 · VBench prompts (Wan2.1) | TCE: 19.76 -> 16.71 | Hurt |
| rewrite prompts with Prompt-A-Video | Wan2.1 · VBench prompts (Wan2.1) | TIE: 39.7 -> 24.10 | Hurt |
| add the DPP diversity term to the relevance-only reward (ablation) | DPP-GRPO · VBench prompts (Wan2.1, reward ablation) | TCE: 20.06 -> 31.95 | Improved |
| add the relevance term to the DPP-only reward (ablation) | DPP-GRPO · VBench prompts (Wan2.1, reward ablation) | CLIP: 0.255 -> 0.311 | Improved |
| add the relevance term to the DPP-only reward (ablation) | DPP-GRPO · VBench prompts (Wan2.1, reward ablation) | TCE: 27.05 -> 31.95 | Improved |
| add GRPO training after supervised fine-tuning (ablation) | DPP-GRPO · VBench prompts (Wan2.1, reward ablation) | TCE: 14.05 -> 31.95 | Improved |
| increase reference set size from 5 to 10 (ablation) | DPP-GRPO · Reference set size (supplement) | TCE: 16.570 -> 15.546 | Hurt |
| raise diversity weight from 0.1 to 0.9 with relevance weight 0.9 to 0.1 (ablation) | DPP-GRPO · Reward weight (supplement) | CLIP: 0.305 -> 0.285 | 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.
- Prompt-A-Video — Prompt-A-Video: Prompt Your Video Diffusion Model via Preference-Aligned LLM (ICCV 2025) · better on 5 of 5 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
- Improving Video Generation with Human Feedback (NeurIPS 2025 · similarity 0.88)
- VidProM: A Million-scale Real Prompt-Gallery Dataset for Text-to-Video Diffusion Models (NeurIPS 2024 · similarity 0.88)
- AdaDiff: Adaptive Step Selection for Fast Diffusion Models (AAAI 2025 · similarity 0.87)
- VideoDPO: Omni-Preference Alignment for Video Diffusion Generation (CVPR 2025 · similarity 0.87)
- Optical-Flow Guided Prompt Optimization for Coherent Video Generation (CVPR 2025 · similarity 0.87)
- EIDT-V: Exploiting Intersections in Diffusion Trajectories for Model-Agnostic, Zero-Shot, Training-Free Text-to-Video Generation (CVPR 2025 · 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.
