Generate updates for LMSYS Chatbot Arena Leaderboard

[zhoukuncheng]

Category

Programming

Website URL

https://chat.lmsys.org/?leaderboard or its mirror in huggingface: https://huggingface.co/spaces/lmsys/chatbot--leaderboard

Website description

LMSYS Chatbot Arena is a crowdsourced open platform for LLM evals. We've collected over 800,000 human pairwise comparisons to rank LLMs with the Bradley-Terry model and display the model ratings in Elo-scale.

What content should be included?

leaderboard for different categories (e.g. overall, coding, long user , English, Chinese)

Additional description

No response

This is not a duplicated rss request

• [X] I have searched existing and pull requests to this rss proposal has not already been requested

[zhoukuncheng]

@TonyRL The original page (https://chat.lmsys.org/?leaderboard) uses Cloudflare to prevent crawling, but their mirror page on Hugging Face (https://huggingface.co/spaces/lmsys/chatbot-arena-leaderboard) doesn't seem to have Cloudflare's reCAPTCHA.

Hugging Face also provides Python/JavaScript APIs, for example:

import { Client } from "@gradio/client";

const app = await Client.connect("lmsys/chatbot-arena-leaderboard");
const result = await app.predict("/update_leaderboard_and_plots", [     
                "Overall", // string  in 'Category' Dropdown component
    ]);

console.log(result.data);

However, I haven't been able to successfully run this code in my environment. It might require some debugging.