slime is an LLM post-training framework for RL Scaling.

J-tau: A Japanese tau-bench for Benchmarking Tool-Agent-User Interaction in Real-World Domains

A fast and soft pattern search for trillion-scale corpora.

MO-GRPO: Mitigating Reward Hacking of Group Relative Policy Optimization on Multi-Objective Problems (Submitted on 26 Sep 2025)

[EMNLP2025] Remedy: Learning Machine Translation Evaluation from Human Preferences with Reward Modeling

Muon is an optimizer for hidden layers in neural networks

Helpful kernel tutorials, examples and SKILLs for tile-based GPU programming

We introduce COMMUNITYNOTES dataset for predicting note helpfulness and its reasons, propose an automatic reason-optimization framework, and show its benefits for evidence sufficiency and fact-chec…

A simple job queue using 'tmux'

Dataset for evaluating the knowledge of Yokai in language models.

Code of "Evaluation of Best-of-N Sampling Strategies for Language Model Alignment"

Code of "Regularized Best-of-N Sampling with Minimum Bayes Risk Objective for Language Model Alignment" (2025).

Evaluate your LLM's response with Prometheus and GPT4 💯

AirLLM 70B inference with single 4GB GPU

Code of Annotation-Efficient Language Model Alignment via Diverse and Representative Response Texts (EMNLP Findings 2025)

A curated list of research papers and resources on Cultural LLM.

Code of "Model-Based Minimum Bayes Risk Decoding for Text Generation" 2024

Flexible evaluation tool for language models

A library for minimum Bayes risk (MBR) decoding

Robust recipes to align language models with human and AI preferences

[EMNLP 2024] Introducing Filtered Direct Preference Optimization (fDPO) that enhances language model alignment with human preferences by discarding lower-quality samples compared to those generated…

The Prism Alignment Project