PURE refines feed-forward 3D reconstruction with unlabeled post-training, big gains on dynamic videos
zhenjun_zhao · x · 2026-10-21
PURE is an annotation-free post-training framework that distills relative depth priors and enforces correspondence-based pose consistency into feed-forward 3D reconstruction models. Applied to VGGT, Pi3, and DepthAnything3, it consistently improves video depth and camera pose estimation on dynamic real-world benchmarks — e.g. VGGT's Sintel Abs Rel drops from 0.279 to 0.220 and ATE from 0.148 to 0.061 — without any ground-truth supervision.
More from Research
- Laude Institute launches FrontierCS 2: agents must beat deleted paper contributions blind — lateinteraction · 2026-10-23
- OPAL: an LLM agent rewrites the optimizer's source code during training — RobertTLange · 2026-10-23
- RFD4-Proteina's 'underrated' docs detail partial diffusion for protein redesign — AllThingsApx · 2026-10-23
- MPI-Sec PhD Thesis on Vulnerability Detection Benchmarks Includes Pre-LLM Classic "Top Score on the Wrong Exam" — mboehme_ · 2026-10-23
- LiveNerf's 29-day data finds no nerfing of Claude Opus 5.5 — TheOnlyVibemaster · 2026-10-23
- New paper makes 9 falsifiable predictions toward a major AI-enabled terror attack by 2027-2028 — StephenLCasper · 2026-10-23