MVPBench

MVPBench is a curated benchmark designed to evaluate visual physical reasoning in multimodal large language models through interleaved multi-image inputs that require coherent, step-by-step reasoning paths.

Papers using MVPBench (10)

Pushupbench: Your VLM Is Not Good At Counting Pushups2026LongVPO: From Anchored Cues to Self-Reasoning for Long-Form Video Preference Optimization2026Clue Matters: Leveraging Latent Visual Clues to Empower Video Reasoning2026MACD: Model-Aware Contrastive Decoding via Counterfactual Data2026Improving Video Question Answering through query-based frame selection2026Video Evidence to Reasoning Efficient Video Understanding via Explicit Evidence Grounding2026Seeing Is Not Reasoning: Mvpbench For Graph-based Evaluation Of Multi-path Visual Physical Cot2025Enhancing Temporal Understanding In Video-llms Through Stacked Temporal Attention In Vision Encoders2025Ro-bench: Large-scale Robustness Evaluation Of Mllms With Text-driven Counterfactual Videos2025Gam-agent: Game-theoretic And Uncertainty-aware Collaboration For Complex Visual Reasoning2025