VBench

'VBench' is a benchmark dataset used to evaluate the performance of video generation models in terms of background consistency and motion continuity in long-horizon nature videos.

Papers using VBench (12)

Do LLMs Favor Their Providers? Measuring Vertical Integration Bias in Code Generation2026Aligning Text, Code, and Vision: A Multi-Objective Reinforcement Learning Framework for Text-to-Visualization2026PISCES: Annotation-free Text-to-Video Post-Training via Optimal Transport-Aligned Rewards2026Light Forcing: Accelerating Autoregressive Video Diffusion via Sparse Attention2026τ-Knowledge: Evaluating Conversational Agents over Unstructured Knowledge2026SpecEyes: Accelerating Agentic Multimodal LLMs via Speculative Perception and Planning2026Causal Forcing++: Scalable Few-Step Autoregressive Diffusion Distillation for Real-Time Interactive Video Generation2026Steady-Forcing: Balancing Spatial Persistence and Motion Continuity in Long-Horizon Nature Video Diffusion2026Patch Validation in Automated Vulnerability Repair2026From Stochastic Answers to Verifiable Reasoning: Interpretable Decision-Making with LLM-Generated Code2026Towards Repository-Level Program Verification with Large Language Models2025How Effective Are Neural Networks for Fixing Security Vulnerabilities2023 · 81 cites