Research Tools: code, datasets, & models

This is the official codebase for the paper “The Belief State Transformer”, based on the nanoGPT repository by Andrej Karpathy.

debug-gym is a text-based interactive debugging framework, designed for debugging Python programs.

In this work, we developed a low-cost, smartphone-based automated Non-Invasive Break-Up Time (NIBUT) measurement prototype. It can be used to estimate dryness of the eye. We utilized the 3D-printed Placido ring attachment on a smartphone’s…

PromptWizard is a self-evolving framework that automates prompt optimization by iteratively refining instructions and in-context examples using feedback from LLMs. It jointly optimizes prompts and examples, incorporates expert reasoning, and adapts dynamically to diverse tasks—delivering…

MatterGen is a generative model for inorganic materials design across the periodic table that can be fine-tuned to steer the generation towards a wide range of property constraints.

HeurAgenix is a novel framework based on LLM, designed to generate, evolve, evaluate, and select heuristic algorithms for solving combinatorial optimization problems. It leverages the power of large language models to autonomously handle various optimization…

MarS is a cutting-edge financial market simulation engine powered by the Large Market Model (LMM), a generative foundation model.

A family of versatile and state-of-the-art video tokenizers.

MageBench is a benchmark for evaluating the reasoning and planning ability of large multimodal model agents. This benchmark currently includes three types of environments: WebUI, Sokoban, and Football, comprising a total of 483 different scenarios.…

Reducio-VAE is a model for encoding videos into an extremely small latent space. It is part of the Reducio-DiT, which is a highly efficient video generation method. Reducio-VAE encodes a 16-frame video clip to T/4∗H/32∗W/32…