Image Captioning on COCO (Coco, Avg %)
100COCO ScorePiToMe
Evaluation results
| Method | Links | ||
|---|---|---|---|
| PiToMeBackbone=InternVL2-8B, FLOPs=22.1, Memory=0.15 GB, TTFT=0.396 sec, Retention Rate=55%2026.05 | 100 | 95.54 | |
| PiToMeBackbone=InternVL2-8B, FLOPs=22.1, Memory=0.15 GB, TTFT=0.396 sec, Retention Rate=60%2026.05 | 99 | 95.99 | |
| LightKVBackbone=InternVL2-8B, FLOPs=23.1, Memory=0.15 GB, TTFT=0.391 sec, Retention Rate=60%2026.05 | 91 | 100.19 | |
| VanillaBackbone=InternVL2-8B, FLOPs=35.7, Memory=0.24 GB, TTFT=0.460 sec, Retention Rate=100%2026.05 | 90 | 100 | |
| LightKVBackbone=InternVL2-8B, FLOPs=23.1, Memory=0.15 GB, TTFT=0.391 sec, Retention Rate=55%2026.05 | 88 | 99.58 | |
| ToMe (P)Backbone=InternVL2-8B, FLOPs=22.1, Memory=0.15 GB, TTFT=0.397 sec, Retention Rate=60%2026.05 | 87 | 97.86 | |
| ToFuBackbone=InternVL2-8B, FLOPs=22.1, Memory=0.15 GB, TTFT=0.395 sec, Retention Rate=60%2026.05 | 81 | 95.49 | |
| ToMe (P)Backbone=InternVL2-8B, FLOPs=22.1, Memory=0.15 GB, TTFT=0.397 sec, Retention Rate=55%2026.05 | 81 | 95.62 | |
| FastVBackbone=InternVL2-8B, FLOPs=24.8, Memory=0.15 GB, TTFT=0.520 sec, Retention Rate=60%2026.05 | 80 | 81.9 | |
| ToFuBackbone=InternVL2-8B, FLOPs=22.1, Memory=0.15 GB, TTFT=0.395 sec, Retention Rate=55%2026.05 | 75 | 95.82 | |
| FastVBackbone=InternVL2-8B, FLOPs=22.9, Memory=0.14 GB, TTFT=0.517 sec, Retention Rate=55%2026.05 | 68 | 79.49 |