Controlled Caption Generation on COCO Localized Narratives (test)
0.665BLEU-1TraceVision-7B
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| TraceVision-7BModel Size=7B2026.02 | 0.665 | 0.328 | 0.276 | 0.492 | 1.53 | 0.328 | |
| Qwen2.5 VL-7BModel Size=7B2026.02 | 0.63 | 0.295 | 0.26 | 0.486 | 1.5 | 0.32 | |
| MITR2026.02 | 0.607 | 0.292 | 0.263 | 0.487 | 1.485 | 0.317 | |
| Ferret-13BModel Size=13B2026.02 | 0.6 | 0.283 | 0.255 | 0.482 | 1.47 | 0.315 | |
| LLaVA 1.5-13BModel Size=13B2026.02 | 0.59 | 0.28 | 0.25 | 0.48 | 1.45 | 0.31 | |
| LN2026.02 | 0.522 | 0.246 | — | 0.483 | 1.065 | 0.365 |