Automatic Speech Recognition on SlideSpeech S95/L95 (dev)
8.3WERSLAM-LLM
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SLAM-LLMTrain set=L95 (473h), Contextual Keywords=true, Adapter=LoRA2026.01 | 8.3 | 5.22 | 8.53 | 94.87 | |
| SLAM-LLMTrain set=L95 (473h), Contextual Keywords=false, Adapter=LoRA2026.01 | 8.82 | 9.62 | 8.77 | 90.38 | |
| SLAM-LLMTrain set=L95 (473h), Contextual Keywords=true2026.01 | 8.91 | 6.07 | 9.13 | 94.02 | |
| SLAM-LLMTrain set=L95 (473h), Contextual Keywords=false2026.01 | 9.38 | 11.98 | 9.19 | 88.08 | |
| SLAM-LLMTrain set=S95 (161h), Contextual Keywords=true2026.01 | 11.14 | 8.92 | 11.36 | 91.44 | |
| SLAM-LLMTrain set=S95 (161h), Contextual Keywords=false2026.01 | 11.57 | 16.23 | 11.28 | 83.83 | |
| LCB-netTrain set=L95 (473h), Contextual Keywords=true2026.01 | 12.21 | 12.12 | 12.21 | 87.98 | |
| CPPTrain set=L95 (473h), Contextual Keywords=true2026.01 | 12.64 | 12.39 | 12.66 | 87.64 | |
| SlidesSpeechTrain set=L95 (473h), Contextual Keywords=false2026.01 | 13.09 | 16.13 | 12.87 | 83.9 | |
| LCB-netTrain set=S95 (161h), Contextual Keywords=true2026.01 | 18.8 | 27.9 | 18.11 | 72.09 | |
| CPPTrain set=S95 (161h), Contextual Keywords=true2026.01 | 20.8 | 28.61 | 20.22 | 71.48 | |
| SlidesSpeechTrain set=S95 (161h), Contextual Keywords=false2026.01 | 21.05 | 31.27 | 20.29 | 68.76 |