THUDM/AlignBench
A comprehensive multi-dimensional benchmark for evaluating Chinese large language model alignment using LLM-as-Judge methodology.
Not currently ranked — collecting fresh signals.
star history
AlignBench is a benchmark designed to evaluate the alignment performance of Chinese large language models. It employs a multi-dimensional, rule-calibrated LLM-as-Judge evaluation approach combined with Chain-of-Thought reasoning to generate analysis and final scores. The benchmark includes a human-involved data construction pipeline to ensure dynamic updates of evaluation data and covers multiple evaluation dimensions to assess real-world model performance.
Frequently asked
- What is THUDM/AlignBench?
- A comprehensive multi-dimensional benchmark for evaluating Chinese large language model alignment using LLM-as-Judge methodology.
- Is AlignBench open source?
- Yes — THUDM/AlignBench is an open-source project tracked on heatdrop.
- What language is AlignBench written in?
- THUDM/AlignBench is primarily written in Python.
- How popular is AlignBench?
- THUDM/AlignBench has 431 stars on GitHub.
- Where can I find AlignBench?
- THUDM/AlignBench is on GitHub at https://github.com/THUDM/AlignBench.