Environment QA on LibriTTS-R clean (test)
52.7Accuracy (Noise)SpeakerLLM-Base
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SpeakerLLM-BaseLLM Backbone=Qwen2.5-1.5B2026.05 | 52.7 | 51.7 | |
| Qwen3.0-Omni-30BLLM Backbone=Qwen3-30B-A3B2026.05 | 32.7 | 20.9 | |
| Audio Flamingo3LLM Backbone=Qwen2.5-7B2026.05 | 21.2 | 21 | |
| Qwen2.5-Omni-7BLLM Backbone=Qwen2.5-7B2026.05 | 20.4 | 20.2 |