Papers › Early-Warning Signals of Grokking via Loss-Landscape Geometry

Early-Warning Signals of Grokking via Loss-Landscape Geometry

19 Feb 2026arXiv:2602.16967added by Syntology

Yongzhong Xu

Title, abstract, authors and date from arXiv's metadata (CC0); this paper is not in the Papers with Code archive (frozen 2025-07-28).

Grokking -- the abrupt transition from memorization to generalization after prolonged training -- has been linked to confinement on low-dimensional execution manifolds in modular arithmetic. Whether this mechanism extends beyond arithmetic remains open. We study two sequence-learning benchmarks: SCAN compositional generalization and Dyck-1 depth prediction. Across both tasks and a wide range of learning rates, the commutator defect -- a curvature measure derived from non-commuting gradient updates -- rises well before generalization, with lead times following a superlinear power law (alpha approximately 1.18 for SCAN, approximately 1.13 for Dyck), consistent with prior results on modular arithmetic. Weight-space PCA reveals that spectral concentration is not a universal precursor; the commutator defect is. Causal interventions demonstrate a mechanistic role: amplifying non-commutativity accelerates grokking (roughly 32% on SCAN, roughly 50% on Dyck), while suppressing orthogonal gradient flow delays or prevents it. The three task families form a spectrum of causal sensitivity -- modular arithmetic is rigid, Dyck is responsive, SCAN is intermediate -- yet suppression delays or prevents grokking in all cases, establishing necessity as a universal finding. These results identify the commutator defect as a robust, architecture-agnostic, causally implicated early-warning signal for delayed generalization in transformers.

PaperPDFCode Syntology ran

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

Syntology 23 code samples harvested from 1 repository linked to this paper: 17 ran (of which 0 constructed an object rather than computing a result; 15 with no instrument failure: 7 honoured, 0 violated, 8 with no contract checked; 2 where Syntology's instrument failed) · 6 unverified. Where Syntology's instrument failed: 1 where our contract draft was wrong; 1 where the contract check could not run; failures of Syntology's instrument, not of the code.

By repository: found in paper text by Syntology: 23 samples from 1 repository, 17 ran (15 with no instrument failure; 2 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

Levels written on 2026-09-26 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

skydancerosel/dyck_scan found in paper text by SyntologySyntology: 23 samples, 17 ran (2 instrument-failed) · 6 unverified report

Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

23 samples harvested: 17 ran (of which 0 constructed an object rather than computing a result; 15 with no instrument failure: 7 honoured, 0 violated, 8 with no contract checked; 2 where Syntology's instrument failed) · 6 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page last changed ; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

7ran · honoured contract
1ran · our draft was wrong
1ran; the contract check could not run
8ran
6unverified

Licence: 23 of the 23 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from skydancerosel/dyck_scan. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity.

skydancerosel/dyck_scan found in paper text by Syntology 23 samples, 17 ran (2 instrument-failed) · 6 unverified
build_depth_dataset dyck/grok_sweep.py ran no licence file found · pointer only · 0cd72a2a0e2529fa · report fixture: signature n_seqs:int/0/int, max_pairs:int/0/int, ctx_len:int/0/int, seed:int/0/int · returned tuple · level recorded 2026-09-26
compute_pc1_from_logs dyck/pc1_lr_experiment.py ran no licence file found · pointer only · f448b03c46a5f7fc · report fixture: signature attn_logs:list/1/dict, subsample:int/0/int · returned list · level recorded 2026-09-26
find_grok_step_acc dyck/pc1_lr_experiment.py ran no licence file found · pointer only · a2fe12212641c856 · report fixture: signature records:list/1/dict, acc_threshold:float/0/float, sustained:int/0/int · returned int · level recorded 2026-09-26
find_grok_step_from_records dyck/generalization_dynamics.py ran · honoured contract no licence file found · pointer only · 708363eac58b21bf · report fixture: signature records:list/1/dict, acc_threshold:float/0/float · returned int · level recorded 2026-09-26
find_grok_step_from_records_acc dyck/lr_sweep.py ran · honoured contract no licence file found · pointer only · 27051bfcfcc000e5 · report fixture: signature records:list/1/dict, acc_threshold:float/0/float · returned int · level recorded 2026-09-26
find_spike_step dyck/generalization_dynamics.py ran no licence file found · pointer only · 034de846e74b7ffb · report fixture: signature records:list/1/dict, threshold_factor:int/0/int, min_defect:int/0/int · returned int · level recorded 2026-09-26
first_acc_cross_step dyck/intervention.py ran · honoured contract no licence file found · pointer only · 96450b853d7a4eb3 · report fixture: signature records:list/1/dict, acc_threshold:float/0/float · returned int · level recorded 2026-09-26
gen_balanced_parentheses dyck/grok_sweep.py ran · honoured contract no licence file found · pointer only · cb70cee49ace8904 · report fixture: signature max_pairs:int/0/int, rng:Random/0/? · returned list · level recorded 2026-09-26
gen_balanced_parentheses dyck/lm_pilot.py ran · honoured contract no licence file found · pointer only · 69efb69bbccec2f6 · report fixture: signature max_pairs:int/0/int, rng:Random/0/? · returned list · level recorded 2026-09-26
hparams_key dyck/intervention.py ran · our draft was wrong no licence file found · pointer only · a83d84e87be0a2dd · report fixture: signature hparams:dict/0/? · returned tuple · level recorded 2026-09-26
make_batch in 20 papers dyck/lm_pilot.py ran no licence file found · pointer only · d2e721359d2c66f3 · report fixture: signature batch_size:int/0/int, ctx_len:int/0/int, max_pairs:int/0/int, device:device/0/?, rng:Random/0/? · returned tuple · level recorded 2026-09-26
masked_ce_loss dyck/lm_pilot.py ran; the contract check could not run (RAISES) no licence file found · pointer only · 026c96222776f842 · report fixture: signature logits:Tensor/3/float, targets:Tensor/2/int · returned Tensor · level recorded 2026-09-26
projected_commutator dyck/commutator_analysis.py ran no licence file found · pointer only · 7fec25438ea4345b · report fixture: signature delta:Tensor/1/float, B:Tensor/2/float, normA:float/0/float, normB:float/0/float, eps:float/0/float · returned dict · level recorded 2026-09-26
random_projection_norm dyck/commutator_analysis.py ran no licence file found · pointer only · 14c07ce05347d233 · report fixture: signature delta:Tensor/2/float, K:int/0/int, n_trials:int/0/int · returned list · level recorded 2026-09-26
split_dataset in 65 papers dyck/grok_sweep.py ran no licence file found · pointer only · afca75c828f1fb23 · report fixture: signature X:Tensor/2/float, Y:Tensor/1/float, frac_train:float/0/float, seed:int/0/int · returned tuple · level recorded 2026-09-26
strict_grok_step dyck/intervention.py ran · honoured contract no licence file found · pointer only · 5c1a373f6061288c · report fixture: signature data:dict/0/? · returned int · level recorded 2026-09-26
strict_grok_step dyck/lr_sweep.py ran · honoured contract no licence file found · pointer only · 0d1db1b7d3cd9c9d · report fixture: signature data:dict/0/? · returned int · level recorded 2026-09-26
collect_trajectory dyck/pca_analysis.py unverified no licence file found · pointer only · b3aabda01158983a · report
compute_per_head dyck/spectral_geometry.py unverified no licence file found · pointer only · 7e7d4efe85773c5e · report
compute_spectral_quantities dyck/spectral_geometry.py unverified no licence file found · pointer only · 93069a8857f32dc6 · report
expanding_window_pca dyck/pca_analysis.py unverified no licence file found · pointer only · a5887c67a2377bf7 · report
load_run in 15 papers dyck/spectral_geometry.py unverified no licence file found · pointer only · 8e8adf55588d8b99 · report
pca_on_trajectory dyck/pca_analysis.py unverified no licence file found · pointer only · 86bf8bb38dae733f · report

Results from the paper

The Papers with Code archive ends with its 2025-07-28 snapshot. This paper's arXiv identifier, 2602.16967, was issued in February 2026, after that date, so the archive has no leaderboard rows for it.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections