Pith. sign in

REVIEW 7 cited by

Grokking at the Edge of Numerical Stability

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2501.04697 v2 pith:RHDJK25A submitted 2025-01-08 cs.LG cs.AIcs.CVstat.ML

Grokking at the Edge of Numerical Stability

classification cs.LG cs.AIcs.CVstat.ML
keywords grokkinggeneralizationregularizationscalingtaskswithoutcontributionsdelayed
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Grokking, the sudden generalization that occurs after prolonged overfitting, is a surprising phenomenon challenging our understanding of deep learning. Although significant progress has been made in understanding grokking, the reasons behind the delayed generalization and its dependence on regularization remain unclear. In this work, we argue that without regularization, grokking tasks push models to the edge of numerical stability, introducing floating point errors in the Softmax function, which we refer to as Softmax Collapse (SC). We demonstrate that SC prevents grokking and that mitigating SC enables grokking without regularization. Investigating the root cause of SC, we find that beyond the point of overfitting, the gradients strongly align with what we call the na\"ive loss minimization (NLM) direction. This component of the gradient does not alter the model's predictions but decreases the loss by scaling the logits, typically by scaling the weights along their current direction. We show that this scaling of the logits explains the delay in generalization characteristic of grokking and eventually leads to SC, halting further learning. To validate our hypotheses, we introduce two key contributions that address the challenges in grokking tasks: StableMax, a new activation function that prevents SC and enables grokking without regularization, and $\perp$Grad, a training algorithm that promotes quick generalization in grokking tasks by preventing NLM altogether. These contributions provide new insights into grokking, elucidating its delayed generalization, reliance on regularization, and the effectiveness of existing grokking-inducing methods. Code for this paper is available at https://github.com/LucasPrietoAl/grokking-at-the-edge-of-numerical-stability.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 7 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Dead-Direction Conditioners: Gauge-Equivariant Preconditioning for Deep Networks

    cs.LG 2026-06 unverdicted novelty 7.0

    Dead-Direction Conditioners provide gauge-equivariant preconditioning by conditioning optimizer state on symmetry orbits, yielding improved resistance to over-training collapse and higher detection of dead directions ...

  2. What Does the Weight Norm Control in Grokking? Logit-Scale Mediation under Cross-Entropy

    cs.LG 2026-06 conditional novelty 7.0

    Grokking delay under cross-entropy is mediated primarily by logit scale and resulting softmax saturation, with weight norm acting only as an upstream handle that adds 1-2% beyond the scale.

  3. Egalitarian Gradient Descent: A Simple Approach to Accelerated Grokking

    cs.LG 2025-10 unverdicted novelty 7.0

    EGD equalizes gradient speeds across singular directions, eliminating or shortening grokking plateaus on modular addition and sparse parity problems.

  4. Less is More: Recursive Reasoning with Tiny Networks

    cs.LG 2025-10 unverdicted novelty 7.0

    TRM with 7M parameters achieves 45% accuracy on ARC-AGI-1 and 8% on ARC-AGI-2, surpassing most LLMs with under 0.01% of their parameters.

  5. Reinforcement Learning for Reasoning in Large Language Models with One Training Example

    cs.LG 2025-04 accept novelty 7.0

    One training example via RLVR boosts LLM math reasoning from 17.6% to 35.7% average across six benchmarks.

  6. Grokking Is Conditional and Fragile: A Fully-Tractable, Multi-Seed Study at 12K Parameters

    cs.LG 2026-07 accept novelty 6.0

    In a fully tractable 12K Llama-style model, grokking is a conditional fragile phase transition gated by coverage (tracking modulus more than structure), weight decay, and floating-point reduction order, so evidence mu...

  7. Distributional Spectral Diagnostics for Localizing Grokking Transitions

    cs.LG 2026-05 unverdicted novelty 6.0

    A residual from Hankel DMD on Wasserstein-mapped training distributions localizes grokking transitions in modular-addition Transformers with AUROC 0.93 and can precede onset under a sustained-threshold rule.