Research map / Vision-language and multimodal / Multimodal LLMs
Leveraging the Power of MLLMs for Gloss-Free Sign Language Translation
Jungeun Kim, Hyeongwoo Jeon, Jongseong Bae, Ha Young Kim
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 MLLM descriptions to images without alignment (ablation) | MMSLT · PHOENIX14T test | BLEU-4: 13.19 -> 12.78 | Hurt |
| add multimodal-language contrastive alignment (ablation) | MMSLT · PHOENIX14T test | BLEU-4: 12.78 -> 25.04 | Improved |
| add GSD-MLLM descriptions on top of alignment (ablation) | MMSLT · PHOENIX14T test | BLEU-4: 23.77 -> 25.04 | Improved |
| add description mapper on top of descriptions + alignment (ablation) | MMSLT · PHOENIX14T test | BLEU-4: 25.04 -> 25.73 | Improved |
| add description mapper without alignment (ablation) | images + descriptions · PHOENIX14T test | BLEU-4: 12.78 -> 24.21 | Improved |
| use LLaVA-OneVision 7B for descriptions instead of InternVL2 8B (ablation) | MMSLT · PHOENIX14T test | BLEU-4: 23.16 -> 25.73 | Improved |
| use detailed prompt (3) instead of simple prompt (2) (ablation) | GSD-MLLM · PHOENIX14T test | BLEU-4: 25.00 -> 25.73 | Improved |
| use in-context prompt (5) instead of detailed prompt (3) (ablation) | GSD-MLLM · PHOENIX14T test | BLEU-4: 25.73 -> 23.10 | Hurt |
| use BERT instead of Qwen2 as description encoder (ablation) | MMSLT · PHOENIX14T test | BLEU-4: 22.64 -> 25.73 | Improved |
| use BERT instead of SentenceBERT as description encoder (ablation) | MMSLT · PHOENIX14T test | BLEU-4: 25.40 -> 25.73 | Improved |
| use MLLM descriptions alone without visual features (ablation) | MMSLT · PHOENIX14T test | BLEU-4: 8.99 vs 25.73 with visual | Hurt |
| MMSLT vs previous best gloss-free (SignLLM) on CSL-Daily | MMSLT · CSL-Daily test | BLEU-4: 15.75 -> 21.11 | Improved |
Closest papers by meaning
- LLMs are Good Sign Language Translators (CVPR 2024 · similarity 0.93)
- Scaling Sign Language Translation (NeurIPS 2024 · similarity 0.92)
- Sign2GPT: Leveraging Large Language Models for Gloss-Free Sign Language Translation (ICLR 2024 · similarity 0.91)
- Improving Gloss-free Sign Language Translation by Reducing Representation Density (NeurIPS 2024 · similarity 0.91)
- Bridging Sign and Spoken Languages: Pseudo Gloss Generation for Sign Language Translation (NeurIPS 2025 · similarity 0.90)
- Conditional Variational Autoencoder for Sign Language Translation with Cross-Modal Alignment (AAAI 2024 · similarity 0.90)
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.
