#video-llm News & Analysis

7 articles tagged with #video-llm. AI-curated summaries with sentiment analysis and key takeaways from 50+ sources.

7 articles

AIBullisharXiv – CS AI · 6d ago7/10

🧠

V-LynX: Token Interface Alignment for Video+X LLMs

Researchers introduce V-LynX, a framework that enhances Video Large Language Models by integrating new sensory modalities through a lightweight auxiliary pathway rather than heavy encoders. The method aligns audio, 3D, and multi-view data with existing video understanding capabilities, achieving state-of-the-art results across multiple benchmarks without requiring paired supervision or freezing the base model.

AINeutralarXiv – CS AI · Apr 157/10

🧠

Distorted or Fabricated? A Survey on Hallucination in Video LLMs

Researchers have conducted a comprehensive survey on hallucinations in Video Large Language Models (Vid-LLMs), identifying two core types—dynamic distortion and content fabrication—and their root causes in temporal representation limitations and insufficient visual grounding. The study reviews evaluation benchmarks, mitigation strategies, and proposes future directions including motion-aware encoders and counterfactual learning to improve reliability.

AINeutralarXiv – CS AI · Mar 177/10

🧠

From Evaluation to Defense: Advancing Safety in Video Large Language Models

Researchers introduced VideoSafetyEval, a benchmark revealing that video-based large language models have 34.2% worse safety performance than image-based models. They developed VideoSafety-R1, a dual-stage framework that achieves 71.1% improvement in safety through alarm token-guided fine-tuning and safety-guided reinforcement learning.

AINeutralarXiv – CS AI · May 116/10

🧠

Response-G1: Explicit Scene Graph Modeling for Proactive Streaming Video Understanding

Response-G1 introduces a novel framework for real-time video understanding that uses explicit scene graphs to align video evidence with query-specific response conditions, enabling Video-LLMs to make more accurate timing decisions during streaming video analysis without requiring fine-tuning.

AINeutralarXiv – CS AI · May 16/10

🧠

Flattery in Motion: Benchmarking and Analyzing Sycophancy in Video-LLMs

Researchers introduce VISE, the first benchmark for evaluating sycophancy in video large language models (Video-LLMs), where models incorrectly agree with user inputs that contradict visual evidence. The study proposes two training-free mitigation strategies: enhanced visual grounding through keyframe selection and inference-time neural representation steering, addressing a critical reliability gap in multimodal AI systems.

AIBearisharXiv – CS AI · Mar 37/108

🧠

VidDoS: Universal Denial-of-Service Attack on Video-based Large Language Models

Researchers have discovered VidDoS, a new universal attack framework that can severely degrade Video-based Large Language Models by causing extreme computational resource exhaustion. The attack increases token generation by over 205x and inference latency by more than 15x, creating critical safety risks in real-world applications like autonomous driving.

AINeutralarXiv – CS AI · Mar 164/10

🧠

Geometry-Guided Camera Motion Understanding in VideoLLMs

Researchers developed a framework to improve video-language models' understanding of camera motion through geometric analysis. The study introduces CameraMotionDataset and CameraMotionVQA benchmark, revealing that current VideoLLMs struggle with camera motion recognition and proposing a lightweight solution using 3D foundation models.