Twinkle AI ORG

High-performance LLM evaluation framework with parallel API calls — up to 17× faster than sequential tools. Supports box, math, and logit-based evaluation.

繁中模型研究與實作的 Lab

視覺化與分析 Twinkle Eval 評估結果

Twinkle Hub — community feedback / feature requests for hub.twinkleai.tw

Official repository for the Twinkle AI Late-Night Study Session. Features hands-on Jupyter notebooks, slides, and code for our book club on "Hands-On Large Language Models"

《Reinforcement Learning from Human Feedback》繁體中文全譯本＋每章互動實驗 | Unofficial zh-TW community translation of the RLHF Book with interactive labs

《The Ultra-Scale Playbook》超大規模訓練實戰手冊繁體中文全譯本＋每章互動實驗 | Unofficial zh-TW community translation with interactive labs

This dataset contains the solutions to the problems on LeetCode in Traditional Chinese.

AiyoDesk 是一款 GPL3.0 免費開源授權的整合軟體，設計宗旨在於「以最低的硬體成本打造專用AI伺服器」，目前已整合 Conda Miniforge、llama.cpp、Open-WebUI、ComfyUI 等軟體，可以在一般家用電腦順暢執行各種量化模型，同時在內網共享 AI 文字交談、影像辨識等功能。

A magical AI playground designed to bring the power of Large Language Models (LLMs) directly to your device.

This repository implements a multi-agent red teaming framework designed to test the safety and robustness of large language models (LLMs). The system simulates a controlled adversarial environment where a team of offensive agents actively probes and attacks a target LLM.

🌐 Twinkle AI Official Website

Twinkle Eval Leaderboard is a visualizer for comparing AI model performance with clear visualizations and tables. Twinkle Eval Leaderboard 是一款用於比較 AI 模型效能的視覺化工具，它以清晰的視覺化圖表和表格呈現。

No description.

從零打造 LLM 互動教材（Twinkle AI）— 原創互動講義，對應 Build a Large Language Model (From Scratch) 各章

No description.

A dataset for instruction fine-tune on coding question-answer task.

No description.

No description.