Text-to-Audio evaluation on RELATE (test)
0.544LCCAQA-Score
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| AQA-ScoreFramework=AQA-Score (LALM-ONLY), Backbone=Qwen-2.5-Omni-7B2026.03 | 0.544 | 0.556 | 0.396 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=Qwen2.5-7B, Audio Encoder=M2D2026.03 | 0.54 | 0.517 | 0.367 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=Qwen3, Audio Encoder=M2D2026.03 | 0.526 | 0.52 | 0.367 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=Qwen2.5-3B, Audio Encoder=MGA2026.03 | 0.504 | 0.477 | 0.334 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=Qwen2.5-3B, Audio Encoder=M2D2026.03 | 0.504 | 0.482 | 0.34 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=Qwen3, Audio Encoder=MGA2026.03 | 0.5 | 0.46 | 0.321 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=Qwen2.5-7B, Audio Encoder=MGA2026.03 | 0.494 | 0.459 | 0.321 | |
| M2D-CLAPFramework=CLAPScore (CLAP-ONLY)2026.03 | 0.492 | 0.457 | 0.322 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=AF3, Audio Encoder=M2D2026.03 | 0.491 | 0.464 | 0.327 | |
| AQA-ScoreFramework=AQA-Score (LALM-ONLY), Backbone=AF3-base2026.03 | 0.475 | 0.508 | 0.357 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=Qwen2.5-3B, Audio Encoder=LAION2026.03 | 0.469 | 0.455 | 0.32 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=Qwen3, Audio Encoder=LAION2026.03 | 0.466 | 0.439 | 0.306 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=Qwen2.5-7B, Audio Encoder=LAION2026.03 | 0.455 | 0.435 | 0.304 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=AF3, Audio Encoder=MGA2026.03 | 0.452 | 0.416 | 0.289 | |
| MGA-CLAPFramework=CLAPScore (CLAP-ONLY)2026.03 | 0.446 | 0.411 | 0.286 | |
| AQA-ScoreFramework=AQA-Score (LALM-ONLY), Backbone=Qwen-2.5-Omni-3B2026.03 | 0.443 | 0.453 | 0.327 | |
| AQA-ScoreFramework=AQA-Score (LALM-ONLY), Backbone=AF3-think2026.03 | 0.435 | 0.474 | 0.33 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=AF3, Audio Encoder=LAION2026.03 | 0.413 | 0.393 | 0.273 | |
| LAION-CLAPFramework=CLAPScore (CLAP-ONLY)2026.03 | 0.404 | 0.386 | 0.268 | |
| FLEURFramework=LALM-ONLY, Backbone=Qwen-2.5-Omni-7B2026.03 | 0.399 | 0.407 | 0.286 | |
| FLEURFramework=LALM-ONLY, Backbone=Qwen3-Omni-Instruct2026.03 | 0.396 | 0.451 | 0.316 | |
| RELATEFramework=Baseline2026.03 | 0.385 | 0.383 | 0.265 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=Qwen2.5-3B, Audio Encoder=MS2026.03 | 0.382 | 0.351 | 0.243 | |
| RELATE w/ CBFramework=Baseline2026.03 | 0.377 | 0.374 | 0.259 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=Qwen3, Audio Encoder=MS2026.03 | 0.372 | 0.329 | 0.227 | |
| FLEURFramework=LALM-ONLY, Backbone=Qwen-2.5-Omni-3B2026.03 | 0.363 | 0.408 | 0.287 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=Qwen2.5-7B, Audio Encoder=MS2026.03 | 0.348 | 0.312 | 0.216 | |
| Raw ScoreFramework=LALM-ONLY, Backbone=Qwen-2.5-Omni-3B2026.03 | 0.337 | 0.341 | 0.272 | |
| Raw ScoreFramework=LALM-ONLY, Backbone=Qwen-2.5-Omni-7B2026.03 | 0.328 | 0.316 | 0.262 | |
| Raw ScoreFramework=LALM-ONLY, Backbone=Qwen3-Omni-Instruct2026.03 | 0.317 | 0.348 | 0.282 | |
| FLEURFramework=LALM-ONLY, Backbone=Qwen3-Omni-Think2026.03 | 0.292 | 0.246 | 0.171 | |
| Raw ScoreFramework=LALM-ONLY, Backbone=Qwen3-Omni-Thinking2026.03 | 0.289 | 0.279 | 0.218 | |
| CAF-ScoreFramework=CAF-Score (CLAP + LALM), Backbone=AF3, Audio Encoder=MS2026.03 | 0.27 | 0.234 | 0.16 | |
| MS-CLAPFramework=CLAPScore (CLAP-ONLY)2026.03 | 0.256 | 0.224 | 0.153 | |
| FLEURFramework=LALM-ONLY, Backbone=AF3-Think2026.03 | 0.163 | 0.163 | 0.111 | |
| FLEURFramework=LALM-ONLY, Backbone=AF3-Base2026.03 | 0.151 | 0.216 | 0.148 | |
| Raw ScoreFramework=LALM-ONLY, Backbone=AF3-Think2026.03 | 0.149 | 0.145 | 0.116 | |
| Raw ScoreFramework=LALM-ONLY, Backbone=AF3-Base2026.03 | 0.12 | 0.111 | 0.092 |