HumanNet is a 1M-hour human-centric video dataset with interaction annotations that enables better vision-language-action model performance than equivalent robot data in a controlled test.
Egoschema: A diagnostic benchmark for very long-form video language understanding
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
citation-role summary
dataset 1
citation-polarity summary
roles
dataset 1polarities
background 1representative citing papers
A question-adaptive greedy frame selector combines SigLIP relevance and DINOv2 coverage under a submodular objective with a text classifier routing to preset trade-offs, yielding accuracy gains on MLVU especially at low frame budgets.
HERMES organizes the KV cache into a hierarchical memory to enable real-time streaming video understanding in MLLMs, achieving 10x faster TTFT and up to 11.4% accuracy gains on streaming benchmarks with 68% fewer tokens.
MVBench is a benchmark of 20 temporal video understanding tasks built by transforming static tasks into dynamic ones, with VideoChat2 outperforming prior MLLMs by over 15%.
EgoCoT-Bench provides 3,172 verifiable QA pairs across perception, anticipation, and reasoning tasks on egocentric videos, revealing that many MLLMs give answer-correct but evidence-inconsistent explanations.
citing papers explorer
-
HumanNet: Scaling Human-centric Video Learning to One Million Hours
HumanNet is a 1M-hour human-centric video dataset with interaction annotations that enables better vision-language-action model performance than equivalent robot data in a controlled test.
-
Adaptive Greedy Frame Selection for Long Video Understanding
A question-adaptive greedy frame selector combines SigLIP relevance and DINOv2 coverage under a submodular objective with a text classifier routing to preset trade-offs, yielding accuracy gains on MLVU especially at low frame budgets.
-
HERMES: KV Cache as Hierarchical Memory for Efficient Streaming Video Understanding
HERMES organizes the KV cache into a hierarchical memory to enable real-time streaming video understanding in MLLMs, achieving 10x faster TTFT and up to 11.4% accuracy gains on streaming benchmarks with 68% fewer tokens.
-
MVBench: A Comprehensive Multi-modal Video Understanding Benchmark
MVBench is a benchmark of 20 temporal video understanding tasks built by transforming static tasks into dynamic ones, with VideoChat2 outperforming prior MLLMs by over 15%.
-
EgoCoT-Bench: Benchmarking Grounded and Verifiable Operation-Centric Chain of Thought Reasoning for MLLMs
EgoCoT-Bench provides 3,172 verifiable QA pairs across perception, anticipation, and reasoning tasks on egocentric videos, revealing that many MLLMs give answer-correct but evidence-inconsistent explanations.
- MACS: Modality-Aware Capacity Scaling for Efficient Multimodal MoE Inference