Skip to content
@tml-epfl

Theory of Machine Learning, EPFL

Popular repositories Loading

  1. llm-adaptive-attacks llm-adaptive-attacks Public

    Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks [ICLR 2025]

    Shell 395 47

  2. understanding-fast-adv-training understanding-fast-adv-training Public

    Understanding and Improving Fast Adversarial Training [NeurIPS 2020]

    Python 96 13

  3. llm-past-tense llm-past-tense Public

    Does Refusal Training in LLMs Generalize to the Past Tense? [ICLR 2025]

    Python 78 12

  4. why-weight-decay why-weight-decay Public

    Why Do We Need Weight Decay in Modern Deep Learning? [NeurIPS 2024]

    Python 73 2

  5. os-harm os-harm Public

    OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents [NeurIPS 2025 Spotlight]

    Jupyter Notebook 72 6

  6. sharpness-vs-generalization sharpness-vs-generalization Public

    A modern look at the relationship between sharpness and generalization [ICML 2023]

    Jupyter Notebook 44 5

Repositories

Showing 10 of 21 repositories
  • eos-overtraining Public

    Code for "(How) Learning Rates Regulate Catastrophic Overtraining", COLM 2026

    tml-epfl/eos-overtraining's past year of commit activity
    Python 0 MIT 0 0 0 Updated Aug 3, 2026
  • sparse-attention-dynamics Public

    The code for "Incremental Learning of Sparse Attention Patterns in Transformers"

    tml-epfl/sparse-attention-dynamics's past year of commit activity
    Jupyter Notebook 0 1 0 0 Updated Jul 28, 2026
  • ATWU Public
    tml-epfl/ATWU's past year of commit activity
    Python 0 MIT 0 0 0 Updated Jul 26, 2026
  • tml-epfl/sub-n-grams-are-stationary's past year of commit activity
    Jupyter Notebook 2 0 0 0 Updated May 19, 2026
  • coconut Public Forked from facebookresearch/coconut

    Training Large Language Model to Reason in a Continuous Latent Space

    tml-epfl/coconut's past year of commit activity
    Jupyter Notebook 1 MIT 192 0 0 Updated Mar 9, 2026
  • softmax Public

    Official code for the paper "Gradient Flow Polarizes Softmax Outputs towards Low-Entropy Solutions"

    tml-epfl/softmax's past year of commit activity
    Jupyter Notebook 1 0 0 0 Updated Mar 9, 2026
  • axlearn-mod Public Forked from apple/axlearn

    AXLearn framework slightly modified to extract the attention sinks metrics.

    tml-epfl/axlearn-mod's past year of commit activity
    Python 0 Apache-2.0 415 0 0 Updated Feb 14, 2026
  • os-harm Public

    OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents [NeurIPS 2025 Spotlight]

    tml-epfl/os-harm's past year of commit activity
    Jupyter Notebook 72 Apache-2.0 6 2 2 Updated Sep 19, 2025
  • learning-parametric-distributions-from-samples-and-preferences Public

    Learning Parametric Distributions from Samples and Preferences [ICML 2025]

    tml-epfl/learning-parametric-distributions-from-samples-and-preferences's past year of commit activity
    Julia 1 MIT 0 0 0 Updated May 24, 2025
  • llm-past-tense Public

    Does Refusal Training in LLMs Generalize to the Past Tense? [ICLR 2025]

    tml-epfl/llm-past-tense's past year of commit activity
    Python 78 12 1 0 Updated Jan 23, 2025