ReQuest introduces an uncertainty-driven question-adaptive keyframe selector with rethinking routing and adaptive NMS that boosts long-form video QA accuracy on Video-MME, MLVU, and LongVideoBench without fine-tuning the base MLLM.
In: Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
years
2026 6representative citing papers
A temperature-perturbed black-box attack infers video training membership in VideoLLMs with 0.68 AUC by exploiting sharper generation behavior on member samples.
Egocentric Scene Graphs convert long videos into short structured text so MLLMs can answer questions about entire sequences, achieving SOTA on HD-EPIC VQA.
SurgOnAir introduces a streaming vision-language model trained on a hierarchical surgical dataset to generate real-time, multi-level narrations with explicit transition tokens.
Sink-Token-aware Pruning (SToP) uses a sink score to suppress attention-sink tokens during visual token pruning, improving fine-grained video understanding in Video LLMs at high pruning rates.
VLMs fail at video emotion recognition because they collapse rare emotions into common ones and process frames as an unordered bag; summarizing skipped frames in text partially recovers temporal cues.
citing papers explorer
-
ReQuest: Rethinking-based Question-Aware Frame Selection for Long-Form Video QA
ReQuest introduces an uncertainty-driven question-adaptive keyframe selector with rethinking routing and adaptive NMS that boosts long-form video QA accuracy on Video-MME, MLVU, and LongVideoBench without fine-tuning the base MLLM.
-
Membership Inference Attacks Against Video Large Language Models
A temperature-perturbed black-box attack infers video training membership in VideoLLMs with 0.68 AUC by exploiting sharper generation behavior on member samples.
-
Graph it first! Enabling Reasoning on Long-form Egocentric Videos through Scene Graphs
Egocentric Scene Graphs convert long videos into short structured text so MLLMs can answer questions about entire sequences, achieving SOTA on HD-EPIC VQA.
-
SurgOnAir: Hierarchy-Aware Real-Time Surgical Video Commentary
SurgOnAir introduces a streaming vision-language model trained on a hierarchical surgical dataset to generate real-time, multi-level narrations with explicit transition tokens.
-
Sink-Token-Aware Pruning for Fine-Grained Video Understanding in Efficient Video LLMs
Sink-Token-aware Pruning (SToP) uses a sink score to suppress attention-sink tokens during visual token pruning, improving fine-grained video understanding in Video LLMs at high pruning rates.
-
Why Do Vision Language Models Struggle To Recognize Human Emotions?
VLMs fail at video emotion recognition because they collapse rare emotions into common ones and process frames as an unordered bag; summarizing skipped frames in text partially recovers temporal cues.