Research Tools: code, datasets, & models

The absolute trainer to light up AI agents. We present Agent Lightning, a flexible and extensible framework that enables seamless agent optimization for any existing agent framework.

AI Behavioral Science for Anthropomorphic Agents: toward human‑centric, symbiotic, and autonomous AI. This project centers on evaluating and promoting anthropomorphic intelligence—AI agents with a human-like mindset and a degree of awareness, capable of acting proactively…

Source code for the paper: Echoes in AI: Quantifying Lack of Plot Diversity in LLM Outputs

TRELLIS is a large 3D asset generation model that creates high-quality 3D assets from simple text or image inputs. Using a unified latent space (SLAT), it delivers detailed, textured 3D models in formats like meshes,…

ChatBench Interactive Benchmark Simulator enables automated, realistic evaluation of AI models through simulated user-AI conversations. We release fine-tuned user simulators (model weights) and supporting infrastructure, allowing the community to assess models in interactive, user-in-the-loop scenarios.

Enhancing the ability of coding models to handle diverse editing requirements.

This repository contains the code and data for SimulatorArena, a framework that enables: (1) benchmarking AI assistants through multi-turn conversations with user simulators, and (2) evaluating the reliability of user simulators as proxies for human…

MatterSim is a deep learning model for accurate and efficient materials simulation and property prediction over a broad range of elements, temperatures and pressures to enable in silico materials design.

This repository contains all scripts for re-producing the results of our paper “Lost in Transmission: When and Why LLMs Fail to Reason Globally”.

ExACT is an approach for teaching AI agents to explore more effectively, enabling them to intelligently navigate their environments, gather valuable information, evaluate options, and identify optimal decision-making and planning strategies.