ZGL injects frozen CLIP text embeddings of VLM-generated motion captions into a DCT Transformer via zero-gated adapters and reports lower MPJPE than pose-only baselines on Human3.6M with transfer to CMUMocap.
Title resolution pending
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
-
Zero-Gated Language-conditioned Human Motion Prediction
ZGL injects frozen CLIP text embeddings of VLM-generated motion captions into a DCT Transformer via zero-gated adapters and reports lower MPJPE than pose-only baselines on Human3.6M with transfer to CMUMocap.