Cover image for LMSYS Chatbot Arena: A New Way for Evaluating Language Models
Promptzone - Commumity
Promptzone - Commumity

Posted on

LMSYS Chatbot Arena: A New Way for Evaluating Language Models


The LMSYS Chatbot Arena is an innovative crowdsourced platform that evaluates the efficacy and alignment of large language models (LLMs) with human preferences. By collecting over 800,000 human pairwise comparisons, the platform ranks models using the Bradley-Terry model and displays their ratings on an Elo scale.

Chatbot Arena Overview

Chatbot Arena's Methodology

The Chatbot Arena leverages a pairwise comparison approach where models are evaluated based on direct comparisons in a variety of tasks. This method not only ensures that the evaluations are robust but also allows the platform to aggregate a significant amount of data to statistically rank the models.

Key Features:

  • Crowdsourced Data Collection: Over 1,007,236 votes have been collected, ensuring a diverse and comprehensive dataset.
  • Bradley-Terry Model: This statistical model provides a reliable method for ranking based on pairwise comparisons.
  • Elo Scale Ratings: Similar to chess rankings, the Elo scale gives a clear, hierarchical ranking of model performance.

Current Leaders and Insights

As of the latest update, the leaderboard is highly competitive with significant participation from major organizations. Notable entries include:

  • GPT-4-Turbo-2024-04-09 by OpenAI, leading with an Elo of 1258 Not surprisingly.
  • Llama-3-70b-Instruct highlighted as the best open-source model currently available.
  • Reka-Core-20240501, a model that is gaining attention for its performance despite less visibility.

Each model's rank, Elo score, confidence intervals, and other pertinent details like organization and licensing are meticulously updated to reflect the latest evaluations.

Model Performance Snapshot

Participate in the Chatbot Arena

The Chatbot Arena is dependent on community participation. By casting your vote at, you contribute to the robust evaluation of future language technologies. The platform is user-friendly and offers a direct way for individuals to impact the development of LLMs.

Future Directions

The Chatbot Arena is not only a tool for ranking LLMs but also a valuable resource for researchers and developers. The ongoing collection of data and the refinement of ranking methodologies continue to enhance our understanding and development of AI models.

For more detailed information and to view the full leaderboard, visit

Top comments (0)