Zeroshot Video Question Answer On Msvd Qa

评估指标

Accuracy

Confidence Score

评测结果

各个模型在此基准测试上的表现结果

			Paper Title
Flash-VStream	80.3	3.9	Flash-VStream: Memory-Based Real-Time Understanding for Long Video Streams
Tarsier (34B)	80.3	4.2	Tarsier: Recipes for Training and Evaluating Large Video Description Models
LinVT-Qwen2-VL (7B)	80.2	4.4	LinVT: Empower Your Image-level Large Language Model to Understand Videos
VILA1.5-40B	80.1	-	VILA: On Pre-training for Visual Language Models
SlowFast-LLaVA-34B	79.9	4.1	SlowFast-LLaVA: A Strong Training-Free Baseline for Video Large Language Models
PLLaVA (34B)	79.9	4.2	PLLaVA : Parameter-free LLaVA Extension from Images to Videos for Video Dense Captioning
IG-VLM-34B	79.6	4.1	An Image Grid Can Be Worth a Video: Zero-shot Video Question Answering Using a VLM
TS-LLaVA-34B	79.4	4.1	TS-LLaVA: Constructing Visual Tokens through Thumbnail-and-Sampling for Training-Free Video Large Language Models
PPLLaVA-7B	77.1	4.0	PPLLaVA: Varied Video Sequence Understanding With Prompt Guidance
Elysium	75.8	3.7	Elysium: Exploring Object-level Perception in Videos via MLLM
MovieChat	75.2	2.9	MovieChat: From Dense Token to Sparse Memory for Long Video Understanding
ST-LLM	74.6	3.9	ST-LLM: Large Language Models Are Effective Temporal Learners
MiniGPT4-video-7B	73.92	-	MiniGPT4-Video: Advancing Multimodal LLMs for Video Understanding with Interleaved Visual-Textual Tokens
Video-LaVIT	73.2	3.9	Video-LaVIT: Unified Video-Language Pre-training with Decoupled Visual-Motional Tokenization
VideoGPT+	72.4	3.6	VideoGPT+: Integrating Image and Video Encoders for Enhanced Video Understanding
LLaVA-Mini	70.9	4.0	LLaVA-Mini: Efficient Image and Video Large Multimodal Models with One Vision Token
Video-LLaVA-7B	70.7	3.9	Video-LLaVA: Learning United Visual Representation by Alignment Before Projection
VideoChat2	70.0	3.9	MVBench: A Comprehensive Multi-modal Video Understanding Benchmark
LLaMA-VID-13B (2 Token)	70.0	3.7	LLaMA-VID: An Image is Worth 2 Tokens in Large Language Models
LLaMA-VID-7B (2 Token)	69.7	3.7	LLaMA-VID: An Image is Worth 2 Tokens in Large Language Models

0 of 28 row(s) selected.