Skip to content

@EleutherAI EleutherAI

Change the repository type filter

All

    Repositories list

    175 repositories

    • lm-evaluation-harness

      Public
      A framework for few-shot evaluation of language models.
      transformerlanguage-modelevaluation-framework
      Python
      •
      MIT License
      •2.7k•10k•480•159•Updated Sep 16, 2025Sep 16, 2025

    • bergson

      Public
      Mapping out the "memory" of neural nets with data attribution
      Python
      •
      MIT License
      •5•26•1•2•Updated Sep 16, 2025Sep 16, 2025

    • elk

      Public
      Keeping language models honest by directly eliciting knowledge encoded in their activations.
      Python
      •
      MIT License
      •33•209•15•10•Updated Sep 15, 2025Sep 15, 2025

    • sparsify

      Public
      Sparsify transformers with SAEs and transcoders
      Python
      •
      MIT License
      •83•621•6•4•Updated Sep 15, 2025Sep 15, 2025

    • delphi

      Public
      Delphi was the home of a temple to Phoebus Apollo, which famously had the inscription, 'Know Thyself.' This library lets language models know themselves through automated interpretability.
      Python
      •
      Apache License 2.0
      •44•211•6•5•Updated Sep 15, 2025Sep 15, 2025

    • djinn

      Public
      Provide a lightweight framework for authoring and validating exploitable verifiable coding problems
      Python
      •0•3•0•0•Updated Sep 15, 2025Sep 15, 2025

    • djinn-problems

      Public
      Problems generated by djinn (exploitably verifiable coding problems)
      0•0•0•0•Updated Sep 11, 2025Sep 11, 2025

    • emergent-misalignment

      Public
      Jupyter Notebook
      •
      MIT License
      •66•1•0•0•Updated Sep 8, 2025Sep 8, 2025

    • website

      Public
      New website for EleutherAI based on Hugo static site generator
      HTML
      •7•5•1•2•Updated Aug 18, 2025Aug 18, 2025

    • deep-ignorance

      Public
      Python
      •1•10•0•0•Updated Aug 12, 2025Aug 12, 2025

    • clt-training

      Public
      Sparsify transformers with cross-layer transcoders
      Python
      •
      MIT License
      •83•15•0•2•Updated Aug 12, 2025Aug 12, 2025

    • tuned-lens

      Public
      Tools for understanding how transformer predictions are built layer-by-layer
      Python
      •
      MIT License
      •59•2•0•0•Updated Aug 7, 2025Aug 7, 2025

    • attribute

      Public
      Python
      •6•10•0•1•Updated Aug 6, 2025Aug 6, 2025

    • attention-probes

      Public
      Linear probes with attention weighting
      Python
      •1•6•0•0•Updated Aug 2, 2025Aug 2, 2025

    • verifiers

      Public
      Verifiers for LLM Reinforcement Learning
      Python
      •
      MIT License
      •335•0•0•0•Updated Jul 31, 2025Jul 31, 2025

    • cookbook

      Public
      Deep learning for dummies. All the practical details and useful utilities that go into working with real models.
      Python
      •
      Apache License 2.0
      •42•813•8•1•Updated Jul 29, 2025Jul 29, 2025

    • gpt-neox

      Public
      An implementation of model parallel autoregressive transformers on GPUs, based on the Megatron and DeepSpeed libraries
      transformerslanguage-modelgpt-3deepspeed-library
      Python
      •
      Apache License 2.0
      •1.1k•7.3k•61•24•Updated Jul 23, 2025Jul 23, 2025

    • SkipTranscoderSAEBench

      Public
      Python
      •0•1•0•0•Updated Jul 22, 2025Jul 22, 2025

    • aria-utils

      Public
      MIDI tokenizers and pre-processing utils.
      Python
      •
      Apache License 2.0
      •3•3•3•1•Updated Jul 21, 2025Jul 21, 2025

    • DeeperSpeed

      Public
      DeepSpeed is a deep learning optimization library that makes distributed training easy, efficient, and effective.
      Python
      •
      Apache License 2.0
      •4.6k•168•0•1•Updated Jul 21, 2025Jul 21, 2025

    • aria-amt

      Public
      Efficient and robust implementation of seq-to-seq automatic piano transcription.
      Python
      •
      Apache License 2.0
      •9•53•0•0•Updated Jul 9, 2025Jul 9, 2025

    • aria

      Public
      Official repository for the paper: Scaling Self-Supervised Representation Learning for Symbolic Piano Performance (ISMIR 2025)
      Python
      •
      Apache License 2.0
      •13•70•0•0•Updated Jul 1, 2025Jul 1, 2025

    • nanoGPT-mup

      Public
      The simplest, fastest repository for training/finetuning medium-sized GPTs.
      Python
      •
      MIT License
      •7.5k•160•1•0•Updated Jun 27, 2025Jun 27, 2025

    • pythia

      Public
      The hub for EleutherAI's work on interpretability and learning dynamics
      Jupyter Notebook
      •
      Apache License 2.0
      •193•2.6k•14•3•Updated Jun 9, 2025Jun 9, 2025

    • truffaldino

      Public
      Investigating goal instability in RL
      Python
      •
      MIT License
      •0•1•0•0•Updated Jun 2, 2025Jun 2, 2025

    • open-r1

      Public
      Fully open reproduction of DeepSeek-R1
      Python
      •
      Apache License 2.0
      •2.4k•4•0•0•Updated May 21, 2025May 21, 2025

    • POSER

      Public
      Poser: Unmasking Alignment Faking LLMs by Manipulating Their Internals
      Python
      •4•2•0•0•Updated May 21, 2025May 21, 2025

    • tyche

      Public
      Precisely estimating the volume of basins in neural net parameter space corresponding to interpretable behaviors
      Jupyter Notebook
      •
      Apache License 2.0
      •0•8•0•2•Updated May 21, 2025May 21, 2025

    • rtopk

      Public
      https://github.com/xiexi51/RTopK PyTorch wrapper
      Cuda
      •
      MIT License
      •0•1•0•0•Updated May 20, 2025May 20, 2025

    • wmdp

      Public
      WMDP is a LLM proxy benchmark for hazardous knowledge in bio, cyber, and chemical security. We also release code for RMU, an unlearning method which reduces LLM performance on WMDP while retaining general capabilities.
      Jupyter Notebook
      •
      MIT License
      •38•0•0•0•Updated May 15, 2025May 15, 2025