Display and submit LLM benchmarks
Create demo spaces for models on Hugging Face
Text-To-Speech (TTS) Evaluation using objective metrics.
Convert Hugging Face models to OpenVINO format
Measure over-refusal in LLMs using OR-Bench
View NSQL Scores for Models
Retrain models for new data at edge devices
View and submit machine learning model evaluations
Compare code model performance on benchmarks
Compare and rank LLMs using benchmark scores
Display leaderboard of language model evaluations
Rank machines based on LLaMA 7B v2 benchmark results
Explore and benchmark visual document retrieval models
The 🌐 Multilingual MMLU Benchmark Leaderboard is a platform designed to evaluate and compare the performance of large language models (LLMs) across multiple languages and tasks. It provides a centralized space for researchers and developers to submit, view, and analyze benchmarks of their models, fostering transparency and innovation in the field of multilingual natural language processing.
• Multilingual Support: Evaluate models across a wide range of languages, enabling a comprehensive understanding of their global capabilities.
• Customizable Benchmarks: Define and submit custom benchmarks tailored to specific languages, tasks, or use cases.
• Real-Time Leaderboard: Access up-to-date rankings of models based on their performance across various metrics.
• Detailed Analytics: Dive into in-depth analysis of model performance, including error distributions, cross-lingual capabilities, and more.
• Community-Driven: Engage with a community of researchers and practitioners, fostering collaboration and knowledge sharing.
• Visualization Tools: Utilize interactive charts and graphs to explore and compare model performance effectively.
What does MMLU stand for?
MMLU stands for Multilingual Model Leaders Universe, a benchmarking framework focused on evaluating the capabilities of multilingual models.
Can I submit my own model's benchmarks?
Yes, the platform allows developers to submit benchmarks for their models, provided they adhere to the submission guidelines and data format requirements.
Is the leaderboard updated in real-time?
The leaderboard is updated periodically to reflect the latest submissions and improvements in model performance. While not real-time, it is refreshed regularly to maintain accuracy.