Pith. sign in

REVIEW 7 cited by

Dynamical versus Bayesian Phase Transitions in a Toy Model of Superposition

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 2310.06301 v1 pith:ACJQ7UNF submitted 2023-10-10 cs.LG cs.AI

classification cs.LGcs.AI
keywords learningbayesianlossphasetransitionscomplexitycriticalgons
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
abstract

We investigate phase transitions in a Toy Model of Superposition (TMS) using Singular Learning Theory (SLT). We derive a closed formula for the theoretical loss and, in the case of two hidden dimensions, discover that regular $k$-gons are critical points. We present supporting theory indicating that the local learning coefficient (a geometric invariant) of these $k$-gons determines phase transitions in the Bayesian posterior as a function of training sample size. We then show empirically that the same $k$-gon critical points also determine the behavior of SGD training. The picture that emerges adds evidence to the conjecture that the SGD learning trajectory is subject to a sequential learning mechanism. Specifically, we find that the learning process in TMS, be it through SGD or Bayesian learning, can be characterized by a journey through parameter space from regions of high loss and low complexity to regions of low loss and high complexity.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 7 Pith papers

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

  1. Emergent Misalignment Recruits a Pre-existing Persona Subspace

    cs.LG 2026-07 conditional novelty 7.0 of 10

    Fine-tuning on narrow bad data recruits a low-rank persona subspace already present in a frozen instruction-tuned model; holding that subspace out of activations prevents broad misalignment, and injecting it into the ...

  2. Influence Dynamics and Stagewise Data Attribution

    cs.LG 2025-10 conditional novelty 7.0 of 10

    Using Bayesian influence functions and singular learning theory, the authors show that a sample's influence on a model varies non-monotonically over training, peaking and flipping sign at phase transitions.

  3. The Kinetics of Training: A Driven-Nucleation Rate Law for Emergence, Plasticity Loss, and Circuit Control in Language Models

    cs.LG 2026-07 conditional novelty 6.0 of 10

    Capability formation in small transformers is claimed to obey a driven-nucleation rate law J = Nνσ(c)e^{−βK} − D, read forward as emergence, backward as plasticity loss, and completed as circuit control.

  4. Local Redundancy: An Information-Theoretic Measure of Plasticity from Synthetic Memorization

    cs.LG 2026-07 conditional novelty 6.0 of 10

    An information-theoretic 'local redundancy' is lower-bounded, via an entropy-cancellation argument, by the expected squared gradient norm on synthetic probe data, and this proxy modestly out-predicts existing plastici...

  5. Spectral-LSH: Sub-Quadratic Prompt Compression via Krylov-Projected Locality-Sensitive Hashing

    cs.AI 2026-06 conditional novelty 6.0 of 10

    At compression ratios above 8x, grouping tokens via a Krylov-projected LSH of an implicit attention kernel preserves LLM output quality far better than block averaging, at a large preprocessing latency cost.

  6. Grokking vs. Learning: Same Features, Different Encodings

    cs.LG 2025-02 conditional novelty 6.0 of 10

    Grokked and steadily trained models learn the same features, but steady training can produce much more compressible models in a parameter regime that grokking does not reach.

  7. You Are What You Eat -- AI Alignment Requires Understanding How Data Shapes Structure and Generalisation

    cs.LG 2025-02 conditional novelty 3.0 of 10

    To align powerful AI, researchers must understand how statistical patterns in training data shape the internal structure of models, because that structure, not eval scores, determines generalization.

Pith tools