Research map / Large language models / Language, safety and interpretability
BoostSLT: Boosting Sign Language Translation via a Plug-and-Play Diffusion-Based Semantic Enhancer
Changzhou Han, Wanlun Ma, Xi Tang, Kun Hu, Sheng Wen, Yang Xiang
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 BoostSLT (EAT-Seg + DSR) to CV-SLT | CV-SLT · PHOENIX-2014T Dev | BLEU-4: 29.10 -> 33.48 | Improved |
| add BoostSLT to TwoStreamNetwork | TwoStreamNetwork · PHOENIX-2014T Test | BLEU-4: 28.95 -> 30.12 | Improved |
| add BoostSLT to GASLT | GASLT · PHOENIX-2014T Test | BLEU-4: 15.62 -> 21.95 | Improved |
| add BoostSLT to MMTLB | MMTLB · Auslan-Daily News | BLEU-4: 2.31 -> 12.46 | Improved |
| add BoostSLT to LLM-based Sign2GPT and LiTFiC | Sign2GPT, LiTFiC · PHOENIX-2014T Test | BLEU-4: Sign2GPT 15.23 -> 18.61; LiTFiC 23.62 -> 25.83 | Improved |
| add BoostSLT to CV-SLT on CSL-Daily | CV-SLT · CSL-Daily Test | BLEU-4: 28.94 -> 32.49 | Improved |
| add BoostSLT to TwoStreamNetwork on CSL-Daily | TwoStreamNetwork · CSL-Daily Test | BLEU-4: 25.79 -> 28.96 | Improved |
| add DSR diffusion reconstruction on top of EAT-Seg (ablation) | GASLT pipeline · PHOENIX-2014T Test | BLEU-4: 17.44 (EAT-Seg + GPT) -> 21.95 (EAT-Seg + DSR) | Improved |
| replace random segmentation with EAT-Seg in the DSR pipeline (ablation) | GASLT pipeline with DSR · PHOENIX-2014T Test | BLEU-4: 16.30 -> 21.95 | Improved |
| use random segmentation alone without post-editing (ablation) | GASLT pipeline · PHOENIX-2014T Test | BLEU-4: 15.62 -> 8.03 | Hurt |
| add GPT autoregressive post-editing to direct concatenation (ablation) | GASLT pipeline · PHOENIX-2014T Test | BLEU-4: 15.62 -> 14.71 | Hurt |
| add fine-tuned LLaMA post-editing to direct concatenation (ablation) | GASLT pipeline · PHOENIX-2014T Test | BLEU-4: 15.62 -> 15.36 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Sign2GPT — Sign2GPT: Leveraging Large Language Models for Gloss-Free Sign Language Translation (ICLR 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Bridging Sign and Spoken Languages: Pseudo Gloss Generation for Sign Language Translation (NeurIPS 2025 · 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 0.92)
- Lost in Translation, Found in Context: Sign Language Translation with Contextual Cues (CVPR 2025 · similarity 0.92)
- Lost in Translation, Found in Embeddings: Sign Language Translation and Alignment (NeurIPS 2026 · similarity 0.92)
- ReSCUE: Re-translation with Sentence Commitment for Unsegmented Long-Form Simultaneous Sign Language Translation (NeurIPS 2026 · similarity 0.91)
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.
