KO
|
EN
gitlite — search
Search
#javascript
#python
#hacktoberfest
#react
#ai
#typescript
#llm
#go
#golang
#android
#machine-learning
#rust
#deep-learning
#linux
CompareBench
★ 9
Open GitHub ↗
ACCV 2026: Submission
Download README (.md)
Explore Similar Repositories
NBL
:
Efficient Large Language Model Inference with Neural Block Linearization
ESP32.LD2410.movements-classification-with-Neural-Networks
:
No description available.
dontmock
:
Presentation for the Droidcon Berlin 2025 'How to write a mock-free app'
ai-agents-hackathon
:
Freepik API code examples
Irancell_internet-Free
:
No description available.
// repository documentation
Was this content helpful?
★ 0
(0 ratings)
Select Rating:
★
★
★
★
★
Submit Feedback
Recent Feedback
×
Download README
Do you want to download the
README.md
file for
CompareBench
?
Download (.md)
# CompareBench **Project Page:** - [CompareBench Project Page](https://caijie0620.github.io/CompareBench/) — overview, method, results, and dataset links --- CompareBench is a benchmark for evaluating **visual comparison reasoning** in vision-language models (VLMs), covering four tasks: quantity, temporal, geometric, and spatial. The benchmark contains 1,200 A-D multiple-choice QA pairs across four sub-benchmarks: - CompareTallyBench: 600 quantity-comparison questions - CompareGeometryBench: 200 geometric-comparison questions - CompareSpatialBench: 100 spatial-comparison questions - CompareTemporalBench: 300 temporal-comparison questions The release is organized with two supporting resources: - [TallyBench (2,000 counting images with QA)](https://huggingface.co/datasets/qiuzhangTiTi/TallyBench) - [OmniCaps (516 historical images + 100 celebrity images + 100 landmark images)](https://huggingface.co/datasets/qiuzhangTiTi/OmniCaps) TallyBench supports the quantity split, OmniCaps supports the temporal split, and the geometry/spatial splits use dedicated A-D annotations. The benchmark dataset is available here: - [CompareBench (1,200 QA pairs)](https://huggingface.co/datasets/qiuzhangTiTi/CompareBench) --- ## Contents - `prompts.yaml`: Standardized instruction templates for TallyBench and all four CompareBench sub-benchmarks. - Benchmark datasets (links above). - Project page source (`index.html`). --- 📌 Paper: *CompareBench: A Benchmark for Visual Comparison Reasoning in Vision–Language Models* (ACCV 2026 submission) 📂 Project page and prompts are maintained in this repository; benchmark datasets are hosted on Hugging Face.
