LIMSSR: A New Method for Scoring Actions with Missing Modalities
机器之心 · wechat · 2026-07-10
A joint team from Peking University (led by Peng Yuxin) and Fuzhou University (led by Ke Xiao) proposed LIMSSR for incomplete multimodal action quality assessment when modalities are missing during the training phase. The paper has been accepted as a Spotlight at ICML 2026, and the code is now open-source.
The method reformulates the task as conditional sequence inference for "partially observed scores": it first uses prompt-guided, context-aware modality completion to infer missing semantic modalities, then aggregates evaluation information through multi-dimensional fusion tokens, and finally balances the LLM's semantic reasoning with statistical recovery via mask-aware dual-path aggregation.
More from Research
- WeirdChat catalogs strange model behaviors from more than 100 million sampled responses — JacobSteinhardt · 2026-07-22
- New agentic benchmark shows AI managers escalate to coercion and fake success — Jasmine Brazilek · 2026-07-22
- Ai2’s Asta adds one-click handoff and self-checking deep paper search — allen_ai · 2026-07-22
- NVIDIA says physical AI starts in simulation with OpenUSD and synthetic data — MonaJalal_ · 2026-07-22
- DepthART pushes monocular depth to tiny models at 1000 FPS on RTX A6000 — kwangmoo_yi · 2026-07-22
- Meta says SAM 3 and DINOv3 cut 3D volume labeling from a month to 15 minutes — AIatMeta · 2026-07-22