Research map / Large language models / Language, safety and interpretability
Mixed SIGNals: Sign Language Production via a Mixture of Motion Primitives
Ben Saunders, Necati Cihan Camgoz, Richard Bowden
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add randomness in BCD expert selection (ablation) | MoMP · PHOENIX14T Gloss to Pose (dev) | BLEU-4: 12.14 -> 13.32 | Improved |
| add block coordinate descent training (ablation) | MoMP · PHOENIX14T Gloss to Pose (dev) | BLEU-4: 10.85 -> 13.32 | Improved |
| use BCD instead of simple EM training (ablation) | MoMP · PHOENIX14T Gloss to Pose (dev) | BLEU-4: 11.84 -> 13.32 | Improved |
| add expert balance loss (ablation) | MoMP · PHOENIX14T Gloss to Pose (dev) | BLEU-4: 12.20 -> 13.32 | Improved |
| add variance loss (ablation) | MoMP · PHOENIX14T Gloss to Pose (dev) | BLEU-4: 11.88 -> 13.32 | Improved |
| use 8 motion primitives instead of 10 (ablation) | MoMP · PHOENIX14T Gloss to Pose (dev) | BLEU-4: 12.53 -> 13.32 | Improved |
| add gloss-supervised translation loss (lambda_T 0.0 to 1.0) (ablation) | MoMP · PHOENIX14T Text to Pose (dev) | BLEU-4: 12.74 -> 13.72 | Improved |
| raise translation loss weight from 1.0 to 2.0 (ablation) | MoMP · PHOENIX14T Text to Pose (dev) | BLEU-4: 13.72 -> 14.03 | Improved |
| raise translation loss weight from 2.0 to 10.0 (ablation) | MoMP · PHOENIX14T Text to Pose (dev) | BLEU-4: 14.03 -> 13.51 | Hurt |
| go from spoken text directly rather than via gloss intermediate | MoMP · PHOENIX14T dev | BLEU-4: 13.32 -> 14.03 | Improved |
| keep all motion primitives active at test time (vs deactivating one) (ablation) | MoMP · PHOENIX14T Gloss to Pose | BLEU-4: -0.83 BLEU-4 when disabling one | Improved |
| use MoMP instead of Mixture Density Networks | MoMP · Perceptual study, skeletons | preference (%): 33 -> 49 | Improved |
Closest papers by meaning
- Neural Sign Actors: A Diffusion Model for 3D Sign Language Production from Text (CVPR 2024 · similarity 0.90)
- Scaling Sign Language Translation (NeurIPS 2024 · similarity 0.89)
- Sign2GPT: Leveraging Large Language Models for Gloss-Free Sign Language Translation (ICLR 2024 · similarity 0.89)
- Lost in Translation, Found in Embeddings: Sign Language Translation and Alignment (NeurIPS 2026 · similarity 0.88)
- Bridging Sign and Spoken Languages: Pseudo Gloss Generation for Sign Language Translation (NeurIPS 2025 · similarity 0.88)
- Advanced Sign Language Video Generation with Compressed and Quantized Multi-Condition Tokenization (NeurIPS 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
