First-Passage Prediction of Grokking Delay: ACalibrated Law under AdamW with Causal Validation

• URL: http://arxiv.org/abs/2605.18845v1
• Date: Wed, 13 May 2026 06:33:56 GMT
• Title: First-Passage Prediction of Grokking Delay: ACalibrated Law under AdamW with Causal Validation
• Abstract summary: We derive a closed-form law T_grok - T_mem = (1 / 2 kappa_LL eta) log(V_mem / V_star), where V_t =theta_t||2 is the norm squared parameter, V_star is an architecture-dependent threshold, and kappa_LL absorbs the AdamW correction to the clean-SGD contraction rate 2 eta.<n> Calibrating (kappa_LL, V_star) on a single hyperlanguage cell predicts grokking delays on 26 held-out runs with MAPE 17.7% over a 41
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: We give the first quantitative prediction of grokking delay under AdamW. Treating the delay as a first-passage time, we derive a closed-form law T_grok - T_mem = (1 / 2 kappa_LL eta lambda) log(V_mem / V_star), where V_t = ||theta_t||^2 is the squared parameter norm, V_star is an architecture-dependent threshold, and kappa_LL absorbs the AdamW correction to the clean-SGD contraction rate 2 eta lambda. Calibrating (kappa_LL, V_star) on a single hyperparameter cell predicts grokking delays on 26 held-out runs with MAPE 17.7% over a 41x delay range; the law generalises to MLPs (MAPE 18.0%, N=34) and degrades to 23.3% on cross-task extension (N=46, 43.5x range), with a structured residual in which V_star / V_mem stays comparatively stable within architecture (CV about 14% on the 1L transformer). First-passage of V_t is necessary but not sufficient. A quantile-margin theorem establishes that positive delay requires both norm separation V_mem > V_post and angular reachability of a threshold alpha_star = arcsin(C / V_T_mem^(1/2)), where C is computable from the empirical NTK feature map and the validation-margin quantile. Calibrating C on modulus p=89 predicts alpha_star = 47.2 degrees at p=97 (observed 47.8 degrees, error 1.3%) as a prior cross-cell prediction. Causal interventions that freeze the norm or remove weight decay at memorisation eliminate grokking (0/6 vs. 3/3 baseline), trapping the angular displacement near 12 degrees. kappa_LL is empirically measured per architecture rather than derived from (beta_1, beta_2, epsilon); within-architecture CV stays at most 15% across four architectures, but values differ by about 2x between architectural variants beyond depth alone. Empirical scope is algorithmic tasks (modular arithmetic, sparse parity) under AdamW; whether the law transfers to natural-language scale models is open.

Related papers

• Architecture-Dependent Causal Transfer of Activation States Across Large Language Models [51.56484100374058]
  Internal language activation states can be transferred causally between different large language models.<n>Injecting projected activations into a target model during generation produces a statistically significant, preregistered causal effect.<n>We conclude that end-to-end activation-state transfer between LLMs, as currently implemented, is architecture-dependent rather than universal.
  arXiv  Detail & Related papers  (2026-08-17T09:53:27Z)
• Phase-Drift Limits and Adaptive Quadrature Readout in Programmable Photonic Processors [2.4050736024350994]
  We study the phase-drift penalty that arises when sine and cosine quadratures are measured sequentially rather than simultaneously.<n>We derive an increment-aware estimator from a local state-space model.<n>For ideal balanced Poisson detection, the Fisher information of each quadrature equals its detected signal-photon number.
  arXiv  Detail & Related papers  (2026-08-03T13:58:30Z)
• No Subspace to Track: Non-Identifiability and Optimizer State in Low-Rank Training [0.0]
  Memory-efficients such as GaLore train large language models by projecting gradients onto a rank-r subspace.<n>We show that beyond a small reproducible core, there is no such object.
  arXiv  Detail & Related papers  (2026-07-07T06:06:04Z)
• Amplifying, Not Learning: Fine-Tuned AI Text Detectors Amplify a Pretrained Direction [51.56484100374058]
  Text detectors amplify a pretrained typicality axis; they do not construct an AI-vs-human boundary.<n>On raw encoders before any task supervision, projecting onto centroid(AI)-centroid(HC3) achieves NYT-vs-HC3 AUROC 0.806/0.944/0.834 across three architectures.<n>On RoBERTa-base, raw projection exceeds fine-tuning; on RoBERTa-base, full fine-tuning reduces discrimination below raw on both fluent-formal populations tested.
  arXiv  Detail & Related papers  (2026-05-20T19:08:38Z)
• Entropy Across the Bridge: Conditional-Marginal Discretization for Flow and Schrödinger Samplers [68.9946633641494]
  Flow matching and Schrdinger bridges define probability, yet their inference grids are usually or inherited from one-endpoint diffusion.<n>We derive a conditional-marginal entropy-rate objective for bridge-aware discretization, separating endpoint-conditioned bridge geometry from marginal flow evolution.<n>On EDM/CIFAR-10, the entropic time-discretization gives the best tested five-step FID (186.3 pm 4.0 versus 200.5 pm 2.9 for linear and 238.0 pm 5.3 for cosine)
  arXiv  Detail & Related papers  (2026-05-15T16:11:10Z)
• Measuring Maximum Activations in Open Large Language Models [60.3514350516308]
  We measure global and layerwise maxima on 27 checkpoints from 8 open families spanning dense, MoE, vision-language, intermediate-training, and instruction-tuned variants.<n>We conclude that maximum activation magnitude is a model property tied to family, architecture, and training stage - not a simple byproduct of size.
  arXiv  Detail & Related papers  (2026-05-15T03:31:51Z)
• The Extrapolation Cliff in On-Policy Distillation of Near-Deterministic Structured Outputs [52.709361620508595]
  ListOPD brings a student to in-domain with an 8B-SFT baseline at one-fifth the parameters.<n>On Amazon Fashion, three pre-registered tests--a fine-grid cliff interval test, and a small-clip cross-prediction--fall within their locked prediction windows--with the small-clip value matching the closed-form prediction below grid resolution.
  arXiv  Detail & Related papers  (2026-05-09T06:48:00Z)
• Kaczmarz Linear Attention [11.650692583508663]
  Linear recurrent models compress the context into a fixed-size state, making the rule that forgets, writes, and edits information a central design problem.<n>Gated DeltaNet (GDN) combines gated state decay with delta-rule residual writes, using a learnable coefficient to balance forgetting and update magnitude.<n>We propose Kaczmarz Linear (KLA), a one-scalar modification of GDN that preserves the state shape, gates, linear recurrence, and chunkwise parallel algorithm.
  arXiv  Detail & Related papers  (2026-05-09T01:07:01Z)
• Normalized Architectures are Natively 4-Bit [49.13186675123547]
  nGPT, an architecture that constrains weights and hidden representations to the unit hypersphere, is inherently more robust to low-precision arithmetic.<n>We validate this approach on both a 1.2B dense model and hybrid (Mamba-Transformer) MoE models of up to 3B/30B parameters.
  arXiv  Detail & Related papers  (2026-05-07T11:54:07Z)
• WARP: A Benchmark for Primal-Dual Warm-Starting of Interior-Point Solvers [0.3058685580689604]
  A growing body of work uses machine learning to predict primal warm-start iterates, reporting reductions of 30-46%.<n>We show that these reported gains rest on an inappropriate evaluation baseline.<n>Experiments establish that the complete primal-dual-barrier state $(x*, *, z*, *)$ reduces IPOPT iterations from 23 to 3.
  arXiv  Detail & Related papers  (2026-05-07T06:19:16Z)
• Adaptive Consensus in LLM Ensembles via Sequential Evidence Accumulation: Automatic Budget Identification and Calibrated Commit Signals [0.0]
  DASE produces a commit-type routing partition that generalises across benchmarks.<n> Adaptive stopping, not injection bandwidth, drives accuracy.<n>Injection-based methods exhibit an inverted-U accuracy-vs-inference trajectory.
  arXiv  Detail & Related papers  (2026-05-05T19:24:10Z)
• The Evaluation Blind Spot: A Stereological Theory of Benchmark Coverage for Large Language Models [0.0]
  Empirically, three independent leaderboards (Open LLM v2, an extended 12-benchmark suite, LiveBench) all have d_eff in [2.86, 4.80] on their competitive frontier.<n>A submodular greedy algorithm with the Nemhauser (1 - 1/e) guarantee finds a stable core of 4 benchmarks.<n>A counterfactual validation across 12 internal benchmarks and 27 categories confirms that the eigenstructure predicts which evaluations are irreplaceable.
  arXiv  Detail & Related papers  (2026-04-15T08:56:58Z)
• The Verification Tax: Fundamental Limits of AI Auditing in the Rare-Error Regime [0.0]
  Most cited calibration result in deep learning -- post-temperature-scaling ECE of 0.012 on CIFAR-100 -- is below the statistical noise floor.<n>We prove this is not a failure of the experiment but a law: the minimax rate for estimating calibration error with model error rate epsilon is Theta((Lepsilon/m)2/3), and no estimator can beat it.
  arXiv  Detail & Related papers  (2026-04-14T16:48:24Z)
• BPDQ: Bit-Plane Decomposition Quantization on a Variable Grid for Large Language Models [56.504879072674015]
  We propose Bit-Plane Decomposition Quantization (BPDQ), which constructs a variable quantization grid via bit-planes and scalar coefficients.<n>BPDQ enables serving Qwen2.5-72B on a single GTX 3090 with 83.85% GSM8K accuracy (vs. 90.83% at 16-bit)
  arXiv  Detail & Related papers  (2026-02-04T02:54:37Z)
• Temporal Zoom Networks: Distance Regression and Continuous Depth for Efficient Action Localization [6.908972852063454]
  Temporal action localization requires both precise boundary detection and computational efficiency.<n>We address this through two complementary innovations: Boundary Distance Regression (BDR) and Adaptive Temporal Refinement (ATR)<n>On THUMOS14, our method achieves 56.5% mAP@0.7 with 151G FLOPs, using 36% fewer FLOPs than ActionFormer++ (55.7% mAP@0.7 at 235G)
  arXiv  Detail & Related papers  (2025-11-06T00:41:54Z)
• Minimax Instrumental Variable Regression and $L_2$ Convergence Guarantees without Identification or Closedness [71.42652863687117]
  We study nonparametric estimation of instrumental variable (IV) regressions. We propose a new penalized minimax estimator that can converge to a fixed IV solution. We derive a strong $L$ error rate for our estimator under lax conditions.
  arXiv  Detail & Related papers  (2023-02-10T18:08:49Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.