Research map / Video / Video-language understanding
CURVE: A Benchmark for Cultural and Multilingual Long Video Reasoning
Darshan Singh, Arsha Nagrani, Kawshik Manikantan, Harman Singh, Dinesh Tewari, Tobias Weyand, Cordelia Schmid, Anelia Angelova, Shachi Dave
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add audio to video input (ablation) | Gemini-2.5-Pro · CURVE (6-locale subset) | accuracy: +4.32% average; zh-TW +8.15%, id-ID +7.09% | Improved |
| increase thinking budget from 128 to 2k tokens (ablation) | Gemini-2.5-Pro · CURVE (6-locale subset) | accuracy: 35.9 -> 45.9 | Improved |
| increase thinking budget beyond 2k tokens (ablation) | Gemini-2.5-Pro · CURVE (6-locale subset) | accuracy | No clear effect |
| increase number of input frames from 1 to 512 (ablation) | Gemini-2.5-Pro · CURVE (6-locale subset) | accuracy | Improved |
| use a larger or frontier model | Gemini-2.5-Pro vs Qwen-2.5-VL · CURVE aggregate | accuracy: 12.75 -> 45.07 | Improved |
| run iterative error isolation up to five iterations (ablation) | Gemini-2.5-Pro error analysis · CURVE (6-locale subset) | error discovery: approx. 22% more errors found; 78 additional reasoning errors | Improved |
| evaluate low-resource locales (ar-EG, ta-IN) vs high-resource | Gemini-2.5-Pro and GPT-5 · CURVE (6-locale subset) | cultural perception errors: 1.4 times more errors | Hurt |
Closest papers by meaning
- MMBench-Video: A Long-Form Multi-Shot Benchmark for Holistic Video Understanding (NeurIPS 2024 · similarity 0.88)
- MMWorld: Towards Multi-discipline Multi-faceted World Model Evaluation in Videos (ICLR 2025 · similarity 0.87)
- Culture In a Frame: C$^3$B as a Comic-Based Benchmark for Multimodal Culturally Awareness (ICLR 2026 · similarity 0.87)
- Video-of-Thought: Step-by-Step Video Reasoning from Perception to Cognition (ICML 2024 · similarity 0.87)
- Look, Remember and Reason: Grounded Reasoning in Videos with Language Models (ICLR 2024 · similarity 0.87)
- TOMATO: Assessing Visual Temporal Reasoning Capabilities in Multimodal Foundation Models (ICLR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
