Research map / Large language models / Language, safety and interpretability
Cueing Without Gapping: Cuer-Independent Cued Speech Recognition Powered by Cross-Cuer Invariant Modeling
Fengji Ma, Chenxing Li, Li Liu
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 text auxiliary modality (ablation) | TACIA · Mandarin Chinese CS | WER: 67.1 -> 58.7 | Improved |
| add pose auxiliary modality (ablation) | TACIA · Mandarin Chinese CS | WER: 67.1 -> 65.5 | Improved |
| combine text and pose auxiliaries (ablation) | TACIA · Mandarin Chinese CS | WER: 58.7 -> 57.3 | Improved |
| replace CTC decoding with LLM-based next token prediction (mBART-Large) (ablation) | TACIA · Mandarin Chinese CS | WER: 68.5 -> 57.3 | Improved |
| scale language model from mBART-Large to mT5-XL (ablation) | TACIA VG-NTP · Mandarin Chinese CS | WER: 57.3 -> 51.5 | Improved |
| scale language model to mT5-XL (ablation) | TACIA VG-NTP · Mandarin Chinese CS | inference time (s): 1.14 -> 3.20 | Hurt |
| mask 30% of pose keypoints during training (ablation) | TACIA · Mandarin Chinese CS | WER: 57.3 -> 57.9 | Hurt |
| use TACIA instead of EcoCued | ACSR · British English CS | CER: 73.6 -> 67.2 | Improved |
Closest papers by meaning
- Unified Speech Recognition: A Single Model for Auditory, Visual, and Audiovisual Inputs (NeurIPS 2024 · similarity 0.86)
- Language Without Borders: A Dataset and Benchmark for Code-Switching Lip Reading (NeurIPS 2024 · similarity 0.85)
- Learning Visual Speech Representations via Cross-Modal Distillation and Joint Face-Lip Modeling (NeurIPS 2026 · similarity 0.84)
- Visual Hallucination Elevates Speech Recognition (AAAI 2024 · similarity 0.84)
- VSRo-200: A Romanian Visual Speech Recognition Dataset for Studying Supervision and Multimodal Robustness (NeurIPS 2026 · similarity 0.84)
- ES3: Evolving Self-Supervised Learning of Robust Audio-Visual Speech Representations (CVPR 2024 · similarity 0.84)
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.
