HERMES: KV Cache as Hierarchical Memory for Efficient Streaming Video Understanding
Open paper
Benchmark
Model
Rank
Results
zero-shot-video-question-answer-on-egoschema-fullset
HERMES (LLaVA-OV-7B, 4K tokens)
#32
Accuracy: 60.29
video-question-answering-on-mvbench
HERMES (Qwen2.5-VL-7B, 4K tokens)
#52
Avg.: 65.53
Rank counts only results with a code link.