UE-MCM fuses a small CLIP4CLIP branch for workflow inconsistency and a large Qwen3-VL branch for fine-grained action errors via a collaboration gate, trained with reweighted cross-entropy, AUC learning, and label-aware adjustment for long-tailed egocentric mistake detection.
Dynamic worlds, dynamic hu- mans: Generating virtual human-scene interaction motion in dynamic scenes.arXiv preprint arXiv:2601.19484, 2026
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Understanding-Enhanced Model Collaboration for Long-Tailed Egocentric Mistake Detection
UE-MCM fuses a small CLIP4CLIP branch for workflow inconsistency and a large Qwen3-VL branch for fine-grained action errors via a collaboration gate, trained with reweighted cross-entropy, AUC learning, and label-aware adjustment for long-tailed egocentric mistake detection.