benchmarks evals
One Spatial Model Keeps 3D and 4D from Crowding Each Other Out
DynThinker constrains the parameters most vulnerable to static-dynamic interference while selectively fusing geometry and motion priors.
Summary
DynThinker constrains the parameters most vulnerable to static-dynamic interference while selectively fusing geometry and motion priors.
Training for static 3D reasoning can weaken understanding of motion, while 4D specialization can cost geometric skill. DynThinker traces much of that interference to particular parameter groups, especially late-layer MLPs. Its decoupled moving average limits drift there, while a fusion module adds geometry and temporal priors selectively. Across ten metrics from nine benchmarks, the authors report the highest average among compared systems and leading open-model scores on four named evaluations. The claim is comparative and benchmark-bound, but it identifies where multi-capability training fights itself.
Why it matters
DynThinker constrains the parameters most vulnerable to static-dynamic interference while selectively fusing geometry and motion priors.
Limits and context
No additional limitation was separately recorded.
Key claims
DynThinker constrains the parameters most vulnerable to static-dynamic interference while selectively fusing geometry and motion priors.
Evidence: source-2026-10-15-007
Sources
- arXiv preprint 2610.20598, original v1arXiv / submitting research authors · primary research
Corrections
: Publication note: This October 15 edition was prepared before its scheduled 9:00 AM America/Detroit reveal, but publication was delayed by a production build failure. The original reporting and image-generation timestamps are preserved.