Pairwise LLM judges (A/B/tie): budget-aware multi-turn packing, position-bias correction, pseudo-label distillation. Generalized from the 4th-place (gold) solution to Kaggle LMSYS Chatbot Arena.
⚔️ [ICLR 2026] Official code of "Search Arena: Analyzing Search-Augmented LLMs".
📊 Daily auto-updated snapshots of all Arena AI (LMSYS Chatbot Arena) leaderboards — LLM, Vision, Code, Video, Image & more. Structured JSON with historical tracking.
[KDD 2026] An automatic, extensible Framework to Evaluate User-Proxy Agents for Human-Likeness. 🌟 Star if you like it!
A simple script to export LMArena.ai conversations to Markdown when the session hits the token limit.
A technical guide and live-tracking repository for the world's top AI models, specialized by coding, reasoning, and multimodal performance.
Rank LLM APIs by cost-effectiveness against Arena ELO scores · 按 Arena ELO 性价比对 LLM API 实时定价排名
A Tampermonkey userscript designed to help you easily download your chat conversations from the LMSYS Arena (arena.ai / lmarena.ai). It exports the chat into a clean Markdown format. Features include an 'Auto Download' mode that scrolls and captures the entire thread, and a 'Manual Select' mode for cherry-picking specific messages.
Fine-tuning and comparing DeBERTa-v3, RoBERTa, and ELECTRA on the Chatbot Arena human-preference task with a shared Siamese architecture.
Export Chatbot Arena conversations to Markdown files to bypass session limits and preserve full chat context including hidden thinking blocks.
30 conversational LLM datasets (~7.7M rows) normalized to one unified schema and published as a single HuggingFace dataset with per-source configs.
A 2026 historical tracker and data analysis tool for the LMSYS Chatbot Arena ELO ratings, including coding and reasoning sub-categories.
Pairwise LLM judges (A/B/tie): budget-aware multi-turn packing, position-bias correction, pseudo-label distillation. Generalized from the 4th-place (gold) solution to Kaggle LMSYS Chatbot Arena.
Pairwise LLM judges (A/B/tie): budget-aware multi-turn packing, position-bias correction, pseudo-label distillation. Generalized from the 4th-place (gold) solution to Kaggle LMSYS Chatbot Arena.
Chatbot Arena 人类偏好偏差挖掘：57K 对战的长度/格式/语气偏差分析,LightGBM 偏好预测 ROC-AUC 0.69
ML ensemble for automated AI prompt scoring - 66.7% accuracy across 17,393 prompts, validated via GPT-4 at 77.3% accuracy and 83.0% F1
Unofficial standalone Android client for Arena (arena.ai)
Pairwise LLM judges (A/B/tie): budget-aware multi-turn packing, position-bias correction, pseudo-label distillation. Generalized from the 4th-place (gold) solution to Kaggle LMSYS Chatbot Arena.
LLM evaluation arena with blind side-by-side comparison, ELO ratings, and category leaderboards. Like LMSYS Chatbot Arena, but self-hosted.
To associate your repository with the chatbot-arena topic, visit your repo's landing page and select "manage topics."