View NSQL Scores for Models
Text-To-Speech (TTS) Evaluation using objective metrics.
Merge machine learning models using a YAML configuration file
Measure execution times of BERT models using WebGPU and WASM
Evaluate reward models for math reasoning
Explain GPU usage for model training
Calculate survival probability based on passenger details
Measure over-refusal in LLMs using OR-Bench
Launch web-based model application
Export Hugging Face models to ONNX
Retrain models for new data at edge devices
Display benchmark results
Open Persian LLM Leaderboard
DuckDB NSQL Leaderboard is a tool designed for model benchmarking, allowing users to view and compare NSQL scores for different models. It provides a centralized platform to evaluate and visualize performance metrics, making it easier to analyze and optimize model effectiveness.
• Real-time Tracking:Monitor NSQL scores as they are updated in real-time.
• Interactive Visualization:Explore performance metrics through graphs and charts.
• Filter and Sort:Easily filter and sort models based on criteria like score, date, or model type.
• Side-by-Side Comparison:Compare multiple models directly to identify strengths and weaknesses.
• Data Export:Export leaderboard data for further analysis or reporting.
What does NSQL stand for?
NSQL stands for Natural Language Query Benchmark, a measure of how well a model understands and processes natural language.
How are NSQL scores calculated?
NSQL scores are calculated by evaluating a model's ability to process and answer natural language queries, often based on standardized benchmarks.
Can I export the leaderboard data for my own analysis?
Yes, the DuckDB NSQL Leaderboard provides options to export data in formats like CSV or Excel for further analysis.