PPaperPicks

Atli Kosson

4 papers at tracked venues · 4 at CORE A* · active 2024–2024

Venues

Frequent coauthors

Martin Jaggi
×4
Bettina Messmer
×2
Dongyang Fan
×1
Alexander Hägele
×1
Elie Bakouch
×1
Loubna Ben Allal
×1
Leandro von Werra
×1

Papers

  1. Analyzing & Reducing the Need for Learning Rate Warmup in GPT Training
    NeurIPS 2024 · Atli Kosson,
    Bettina Messmer
    ,
    Martin Jaggi
  2. Ghost Noise for Regularizing Deep Neural Networks
    AAAI 2024 · Atli Kosson,
    Dongyang Fan
    ,
    Martin Jaggi
  3. Rotational Equilibrium: How Weight Decay Balances Learning Across Neural Networks
    ICML 2024 · Atli Kosson,
    Bettina Messmer
    ,
    Martin Jaggi
  4. Scaling Laws and Compute-Optimal Training Beyond Fixed Training Durations
    NeurIPS 2024 ·
    Alexander Hägele
    ,
    Elie Bakouch
    , Atli Kosson,
    Loubna Ben Allal
    ,
    Leandro von Werra
    ,
    Martin Jaggi