Suggestion Generation on L2-Arctic-plus (test)
20.4BLEU-2Instruction-Tuned ALM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Instruction-Tuned ALMAudio Encoder=Whisper Medium, LLM=Llama-3.1-8B2026.01 | 20.4 | 31.9 | 87.4 | |
| Instruction-Tuned ALMAudio Encoder=Whisper Large, LLM=Mistral-7B2026.01 | 20.1 | 30.5 | 87.2 | |
| Instruction-Tuned ALMAudio Encoder=Whisper Large, LLM=Llama-3.1-8B2026.01 | 20 | 30.5 | 87.3 | |
| Instruction-Tuned ALMAudio Encoder=Whisper Medium, LLM=Mistral-7B2026.01 | 19.7 | 30.7 | 87.2 | |
| Instruction-Tuned ALMAudio Encoder=Whisper Small, LLM=Mistral-7B2026.01 | 17.4 | 25.9 | 85.7 | |
| Instruction-Tuned ALMAudio Encoder=Whisper Small, LLM=Llama-3.1-8B2026.01 | 17.2 | 25.4 | 85.5 | |
| GPT-4o-Audio2026.01 | 10.9 | 22.3 | 86 | |
| ASR+LLM Cascade SOTAAudio Encoder=Wav2vec2 Base, LLM=Llama-3.1-8B2026.01 | 7.3 | 15 | 83 |