AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

Not Only Vision: Evolve Visual Speech Recognition via Peripheral Information

Zhaoxin Yuan, Shuang Yang, Shiguang Shan, Xilin Chen

ICCV 2025 · Vision-language and multimodal › Cross-modal learning · approach: Audio-visual speech separation · 2 citations (OpenAlex)

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 changedWhereMeasuredEffect
add all peripheral information (guidance, expertise, perturbation)VSR with Llama3.1-8B · LRS3WER: 24.92 -> 22.03Improved
add contextual guidance (title and description) (ablation)Synergy LoRA · LRS3WER: 25.03 -> 23.46Improved
add all contextual guidance types incl. speaker description (ablation)Synergy LoRA · LRS3WER: 25.03 -> 22.99Improved
add task expertise (Type B) without guidance (ablation)Synergy LoRA · LRS3WER: 25.03 -> 24.44Improved
add task expertise (Type B) on top of guidance (ablation)Synergy LoRA · LRS3WER: 23.46 -> 23.09Improved
add task-aware prompt to task expertise (ablation)Synergy LoRA Type A · LRS3WER: 24.83 -> 24.62Improved
add output constraint to task expertise (ablation)Synergy LoRA Type A · LRS3WER: 24.91 -> 24.62Improved
add random-token linguistic perturbation (ablation)Synergy LoRA · LRS3WER: 25.03 -> 24.62Improved
add natural text clip perturbation (ablation)Synergy LoRA · LRS3WER: 25.03 -> 24.65Improved
add random perturbation to guidance training (ablation)Synergy LoRA · LRS3WER: 23.46 -> 23.12Improved
use Synergy LoRA instead of vanilla LoRA with perturbed guidance (ablation)LoRA · LRS3WER: 24.05 -> 23.12Improved
use Synergy LoRA design (e) instead of vanilla LoRA (a) (ablation)LoRA · LRS3 with peripheral informationWER: 23.92 -> 22.03Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy