SpecWM: Training-Free Refinement Boosts Streaming World Models 1.63x
Huster · hf · 2026-11-20
Researchers from HUST, Tencent, PKU and HKUST present SpecWM, a training-free inference framework for streaming video world models that turns draft verification into refinement.
- Instead of score-only accept/reject, it probes the target model's latent velocity field to produce both a score and a reusable correction for each draft chunk.
- Threshold-based accept/refine/reject routing enables 1–3 refinement steps at pixel and latent levels, with corrections reused across chunks.
- On WBench Navigation it achieves 1.63x speedup over Target-only while improving aesthetics and retaining target-level navigation control.
More from Multimodal
- Google ships 8 updates in 2 weeks: Gemini 4 Neon/Hydrogen/Carbon, Genie 4, Robotics 2.5 — GoogleAI · 2026-11-21
- CapCut plugin connects ChatGPT to auto-cut gameplay highlights, then human picks the payoff — iamfakhrealam · 2026-11-21
- Posterior Matching Distillation preserves teacher diversity in few-step generation — ArashVahdat · 2026-11-21
- Qwen3 TTS goes live on Together AI: 9 voices, plain-language control, 10 languages — togethercompute · 2026-11-21
- 3DGS vs Point Cloud comparison looks great; surface reconstruction for sim collisions next — jonstephens85 · 2026-11-21
- AI video showing 25+ celebrities 'before and after' plastic surgery goes viral — evisoft · 2026-11-21