Multimodal Evaluation on MME (C, P, Avg %)
357.5Conceptual Score (C)LightKV
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| LightKVBackbone=LLaVA-v1.5-7B, Compression Timing=During prefill, FLOPs (Tera)=6.6, Memory (GB)=0.23, TTFT (sec)=0.065, Vision Token Retention=55%2026.05 | 357.5 | 1,519.8 | 99.79 | |
| LightKVBackbone=LLaVA-NeXT-7B, Compression Timing=During prefill, FLOPs (Tera)=22.3, Memory (GB)=0.67, TTFT (sec)=0.259, Vision Token Retention=55%2026.05 | 338.6 | 1,517.3 | 98.85 | |
| LightKVBackbone=LLaVA-NeXT-13B, Compression Timing=During prefill, FLOPs (Tera)=37.3, Memory (GB)=1.05, TTFT (sec)=0.383, Vision Token Retention=55%2026.05 | 326.1 | 1,576.5 | 98.12 | |
| LightKVBackbone=LLaVA-v1.5-13B, Compression Timing=During prefill, FLOPs (Tera)=12.6, Memory (GB)=0.37, TTFT (sec)=0.098, Vision Token Retention=55%2026.05 | 302.1 | 1,543.8 | 99.94 |