Find open-source science resources

Neural network-based cryo-EM heterogeneous reconstruction, modeling continuous 3D structure distributions from single-particle images, with CryoDRGN-ET extending to in-cell cryo-electron tomography (MIT CSAIL, Nature Methods 2021/2024)

Python package for simulation-based inference enabling likelihood-free Bayesian parameter estimation from scientific simulators, with flexible interfaces for neural posterior estimation, sequential methods, and MCMC/variational backends (Mackelab, 825+ stars)

Efficient differentiable n-dimensional PDE solvers built on JAX and Equinox, shipping 46+ built-in equations with Fourier spectral methods, exponential time differencing, and full auto-differentiation for physics-based deep learning workflows (MIT, 200+ stars, 2024)

Differentiable PDE solving framework for machine learning with built-in fluid simulation, supporting PyTorch/JAX/TensorFlow backends and enabling neural network training within physical simulations (TUM, MIT License)

Geometry Aware Operator Transformer serving as an efficient and accurate neural surrogate for PDEs on arbitrary domains, combining geometric priors with transformer architectures for scientific computing (ETH Zurich CAMLab, 92+ stars)

Efficient foundation models for PDEs with pretrained transformer-based neural operators and downstream task fine-tuning pipelines, HuggingFace integration for models and datasets (ETH Zurich CAMLab, arXiv 2024)

Learning operators in Fourier space

Kolmogorov-Arnold Networks with learnable activation functions on edges instead of fixed node activations, achieving strong performance in function fitting, PDE solving, and scientific discovery with enhanced interpretability as an alternative to MLPs (MIT, 16.3K+ stars, 2024)

Parallel symbolic regression network evaluating millions of expressions on GPU with automated subtree reuse, Nature Computational Science cover article (MIT, 2026)

Scientific equation discovery and symbolic regression using LLMs, combining code generation with evolutionary search (ICLR 2025 Oral)

High-performance symbolic regression for discovering interpretable scientific equations from data, multi-population evolutionary search with Python/Julia backend, widely used in physics and astronomy (Cambridge, NeurIPS 2023)

Sparse identification of nonlinear dynamics

Learning nonlinear operators

Physics-informed neural networks in Julia

Keras-based scientific neural networks

Physics-Informed Neural networks for Advanced modeling in PyTorch

Physics-informed neural networks

Deep learning library for solving PDEs

Neural differential equations in Julia

Numerical differential equation solving in JAX

Neural differential equations in PyTorch

PyTorch implementation of neural ODEs

AI agent for therapeutic reasoning across a universe of tools, achieving 92.1% accuracy in drug reasoning and outperforming GPT-4o by 25.8% (Harvard MIMS, 2025)

Bioinspired multi-agent intelligent graph reasoning system that autonomously traverses ontological knowledge graphs to generate, critique, and refine novel research hypotheses, demonstrated on bio-inspired materials discovery with cross-disciplinary connection mining (MIT Lamm Group, 2024)

Large Language Models for automated open-domain scientific hypotheses discovery (ACL 2024, ICML Best Poster)

Fully autonomous medical image segmentation research system that generates complete manuscripts end-to-end from datasets with zero human intervention, beating strongest baselines on 24 of 31 datasets and achieving T1-T2 tier manuscript quality in double-blind evaluations (USTC & Shanghai AI Lab, 2026)

Multimodal LLM-based AI agent enabling deep research in spatial transcriptomics, automating analysis and interpretation of spatial gene expression data (Harvard LiuLab, bioRxiv 2025)

General-purpose biomedical AI agent integrating LLM reasoning with retrieval-augmented planning and code-based execution to autonomously execute diverse biomedical research tasks and generate testable hypotheses (Stanford SNAP, bioRxiv 2025)

AI agent for biological discovery and research automation

LLMs as copilots for theorem proving in Lean 4, exposing native tactics (`suggest_tactics`, `search_proof`, `select_premises`) that embed language model inference and premise retrieval directly inside the Lean proof environment, supporting local CTranslate2/CUDA inference as well as remote model APIs for interactive and automated proof search (Caltech & NVIDIA, NeurIPS 2024, 1.2K+ stars)

Open-source toolkit and benchmark for learning-based theorem proving in Lean, providing programmatic Lean interaction, a 98K+ theorem dataset extracted from 217 Lean projects, and ReProver—the first retrieval-augmented LLM-based theorem prover for Lean—with reproducible training pipelines underpinning much subsequent Lean prover research (Caltech & NVIDIA, NeurIPS 2023 Outstanding Paper, Datasets & Benchmarks)

DeepSeek's open-source large language model for formal theorem proving in Lean 4, integrating informal and formal mathematical reasoning through recursive subgoal decomposition and reinforcement learning powered by DeepSeek-V3, with open weights and ProverBench evaluation (2025)

Strongest open-source automated theorem prover in Lean 4, 8B model matches DeepSeek-Prover-V2-671B at 84.6% MiniF2F, 32B model achieves 90.4% with self-correction, using scaffolded data synthesis and verifier-guided proof refinement (Princeton, 2025)

Web application for LLM-assisted manuscript review and annotation

First benchmark evaluating LLMs' ability to rediscover scientific laws through interactive experimentation across 324 tasks in 12 physics domains, featuring memorization-resistant metaphysical shifts of canonical laws (HKUST)

Research coding benchmark curated by scientists with 338 subproblems across 16 subdomains (physics, math, materials, biology, chemistry), evaluating LLMs on realistic scientific programming tasks with gold-standard solutions (NeurIPS 2024)

First physics-aligned interactive benchmark for LLM agents in engineering construction, designing rockets/cars/bridges in physics simulator with 3D spatial geometry library

Evaluating multimodal autonomous agents in realistic scientific workflows across real scientific software environments (KAlgebra, Celestia, Grass GIS, Lean 4, etc.) with VM-based evaluation infrastructure and agent trajectories

Benchmark evaluating AI agents on 75 curated Kaggle-style ML engineering competitions with reproducible Docker-based grading harness, human baselines, and end-to-end task lifecycle, used as a primary benchmark for autonomous ML research agents (e.g., InternAgent #1 at 36.44%)

Benchmark evaluating AI agents' ability to replicate 20 ICML 2024 Spotlight/Oral papers from scratch, with 8,316 gradable tasks and author-co-developed rubrics

Benchmark quantifying end-to-end autonomous AI research abilities of LLM agents across 20 tasks from SOTA machine learning papers spanning NLP, code, math, biochemical modelling, and time series forecasting, with normalized score metrics against human SOTA and HuggingFace dataset

102 executable tasks from 44 peer-reviewed papers across 4 disciplines with containerized evaluation

Evolvable and privacy-preserving multi-agent framework automating, scaling, and accelerating data sciences with a particular focus on end-to-end single-cell biology analyses; features agentic code evolution, multi-agent team orchestration, distributed architecture, and a community marketplace with 1,000+ curated agents and skills (428+ stars)

Self-evolving AI scientist with 6 specialized sub-agents (plan/research/code/debug/analyze/write) and persistent memory, #1 on DeepResearch Bench II and AstaBench, supporting multi-provider LLMs and multi-channel deployment (Apache 2.0, 2026)

Universal scientific research intelligence covering 50+ disciplines, repositioning LLMs as cross-disciplinary generators with human experts as verifiers; 30B model outperforms Claude Opus and GPT on 5 research benchmarks

Andrej Karpathy's autonomous LLM research framework: AI agent runs overnight experiments on a real training setup, auto-editing code→5min training→evaluation in a loop, ~100 experiments per night on a single GPU

Automated hypothesis testing with agentic sequential falsifications

Automated and rigorous experiments using AI agents for scientific discovery

Language agent gymnasium for challenging scientific tasks including DNA manipulation, literature search, and protein engineering

FutureHouse's end-to-end scientific discovery multi-agent system orchestrating literature search (Crow/Falcon) and data analysis (Finch) agents, first AI-generated drug discovery identifying ripasudil as novel dry AMD therapeutic (2025)