[NeurIPS 2025 D&B Spotlight] Scaling Data for SWE-agents

Harbor is a framework for running agent evaluations and creating and using RL environments.

Entire is a new developer platform that hooks into your git workflow to capture AI agent sessions on every push, unifying your code with its context and reasoning.

Open-source framework for the research and development of foundation models.

SWE-bench: Can Language Models Resolve Real-world Github Issues?

Sandboxed code execution for AI agents, locally or on the cloud. Massively parallel, easy to extend. Powering SWE-agent and more.

The 100 line AI agent that solves GitHub issues or helps you in your command line. Radically simple, no huge configs, no giant monorepo—but scores >74% on SWE-bench verified!

Post-training with Tinker

SWE-agent takes a GitHub issue and tries to automatically fix it, using your LM of choice. It can also be employed for offensive cybersecurity or competitive coding challenges. [NeurIPS 2024]

Minimal, lightweight JAX implementations of popular models.

Legible, Scalable, Reproducible Foundation Models with Named Tensors and Jax

Rust grammar for tree-sitter

Accelerate, Optimize performance with streamlined training and serving options with JAX.

[CVPR 2025 (Oral)] Open implementation of "RandAR"

[ICLR2025] Halton Scheduler for Masked Generative Image Transformer

Pix2Seq codebase: multi-tasks with generative modeling (autoregressive and diffusion)

PyTorch implementation of the Recurrent Interface Network (RIN), based on the original Tensorflow implementation from the pix2seq repository.

This repo contains the code for 1D tokenizer and generator

A high-throughput and memory-efficient inference and serving engine for LLMs