Audio Captioning on Clotho (evaluation)
39CIDEr-DSLAM–AAC
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| SLAM–AACVariant=LoRA, Enc.=✓, LoRA=✓, Distill=–2026.07 | 39 | 25.8 | 12.7 | 16.2 | |
| SLAM–AACVariant=No LoRA, Enc.=✓, LoRA=✗, Distill=–2026.07 | 32 | 22.3 | 12.2 | 16.4 | |
| CARDVariant=CARD∗, Enc.=✗, LoRA=✓, Distill=Projector (early stages) + LLM2026.07 | 27.5 | 18.8 | 10.1 | 14.2 | |
| CARDVariant=CARD⋄, Enc.=✗, LoRA=✓, Distill=Projector (all stages) + LLM2026.07 | 24.8 | 17.1 | 9.4 | 13.1 | |
| CARDVariant=Proj_Early, Enc.=✗, LoRA=✓, Distill=Early stages→Projector2026.07 | 24.3 | 16.6 | 8.9 | 13.1 | |
| CARDVariant=LLM Distill, Enc.=✗, LoRA=✓, Distill=LLM only2026.07 | 22.5 | 15.1 | 8.4 | 13 | |
| CARDVariant=No Distill, Enc.=✗, LoRA=✓, Distill=No teacher2026.07 | 22.3 | 15.5 | 8.6 | 13 | |
| CARDVariant=Proj_Full, Enc.=✗, LoRA=✓, Distill=All stages→projector2026.07 | 21.2 | 14.6 | 8.1 | 12.7 |