Official Implementation of "Maximum Likelihood Reinforcement Learning (MaxRL)"

Solve expert level Sudoku using only logical techniques (in other words no brute forcing or guessing). Outputs a detailed description of the techniques and moves required at each step. In-depth ana…

The rule-based evaluation subset and code implementation of Omni-MATH

A simple yet powerful tool to turn traditional container/OS images into unprivileged sandboxes.

Kimina Lean server (+ client SDK)

Flash Attention Triton kernel with support for second-order derivatives

Optimize prompts, code, and more with AI-powered Reflective Optimization

SkyRL: A Modular Full-stack RL Library for LLMs

verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework

Provide with pre-build flash-attention 2 and 3 package wheels on Linux and Windows using GitHub Actions

Ternary plots as projections of Matplotlib

Save matplotlib figures as TikZ/PGFplots for smooth integration into LaTeX.

LaTeX package for automatically putting proof environments in appendix

Tape Reader (v1.0)

Python package for *fast* TDigest calculation.

A fast t-digest library for Python built on Rust.

Train transformer language models with reinforcement learning.

Efficient Knowledge Injection in LLMs via Self-Distillation (TMLR)

A Lean tactic for Canonical, a search procedure for terms in dependent type theory.

The repository provides code for running inference and finetuning with the Meta Segment Anything Model 3 (SAM 3), links for downloading the trained model checkpoints, and example notebooks that sho…

[EMNLP, Findings 2024] a radiology report generation metric that leverages the natural language understanding of language models to identify and explain clinically significant errors in candidate r…

Official implementation for PedCLIP: A Vision-Language model for Pediatric X-rays with Mixture of Body part Experts [MICCAI 2025]

Code for TFG: Unified Training-Free Guidance for Diffusion Models

Qwen3-VL is the multimodal large language model series developed by Qwen team, Alibaba Cloud.

Improving Performance, Robustness, and Fairness of Radiographic AI Models with Finely-Controllable Synthetic Data

Repository to train Latent Diffusion Models on Chest X-ray data (MIMIC-CXR) using MONAI Generative Models

A large-scale collection of chest X-ray datasets.