The Hidden Ratio in Adam: Stable Structure, Compression, and Sign Dynamics

arXiv:2609.35392 · cs.LG, cs.AI · Submitted 2026-09-28 · Read on arXiv

cs.LG, cs.AI

Submitted: 2026-09-28

Updated: 2026-09-28

Code: https://github.com/Niccolo-Ajroldi/plainLM2https:

Terminology

• Adam — 1× in this paper · explained in 2 episodes

Sources

• Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
• The Effect of Mini-Batch Noise on the Implicit Bias of Adam
• UltraFeedback: Boosting Language Models with Scaled AI Feedback
• Why beta 1 = beta 2 Is Dynamically Special in Adam
• OLMo: Accelerating the Science of Language Models
• Skywork-Reward-V2: Scaling Preference Data Curation via Human-AI Synergy
• LaProp: Separating Momentum and Adaptivity in Adam
• Reuse, Don't Retrain: A Recipe for Continued Pretraining of Language Models
• The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale
• FP8-LM: Training FP8 Large Language Models
• LLaMA: Open and Efficient Foundation Language Models

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks