Scaling Laws for Behavioral Foundation Models over User Event Sequences

• URL: http://arxiv.org/abs/2606.05257v1
• Date: Wed, 03 Jun 2026 15:59:25 GMT
• Title: Scaling Laws for Behavioral Foundation Models over User Event Sequences
• Abstract summary: We study a common two-part behavioral-model architecture, a feature-based event embedder, and a decoder-only transformer.<n>Across roughly 600 runs on real interaction data, spanning $1015$-$1019$ training FLOPs, we vary four deployment-relevant axes.<n> Compute-optimal training is data-heavy relative to text at low compute, but its $D/N$ ratio moves toward the Chinchilla as compute increases.
• Score: 2.924581427482972
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Foundation models are increasingly trained on sequences of user actions in recommendation, payments, fraud, and commerce, but these models still lack the kind of compute calibration that scaling laws provide for language models. We study a common two-part behavioral-model architecture: a feature-based event embedder maps each multi-modal item to a vector, and a decoder-only transformer predicts the next event from the resulting sequence. Across roughly 600 runs on real interaction data, spanning $10^{15}$-$10^{19}$ training FLOPs, we jointly vary four deployment-relevant axes: the two-part parameter split, critical batch size, model/data allocation, and the number of sampled negatives used after freezing the embedder. A small embedder ($s^{\star}\!\approx\!2\%$ of parameters) is compute-optimal at every budget we test because embedder parameters are both more expensive per step and exposed to far more repeated items than contextualizer parameters. Compute-optimal training is data-heavy relative to text at low compute, but its $D/N$ ratio moves toward the Chinchilla heuristic as compute increases. The sampled training objective and deployed ranking metrics disagree in ways that themselves scale: critical batch size, optimal negative count after freezing, and the agreement between loss and ranking quality all shift with compute and with the chosen evaluation metric. For negative sampling, larger budgets increasingly prefer more negatives; by $10^{19}$ FLOPs the active constraint is candidate-axis memory rather than FLOPs. In behavioral foundation models, the evaluation metric is therefore part of the scaling law: changing it can change the compute-optimal recipe.

Related papers

• Test-Time Scaling Makes Overtraining Compute-Optimal [23.520624926542755]
  We present Train-to-Test ($T2$) scaling laws that jointly optimize model size, training tokens, and number of inference samples.<n>$T2$ modernizes pretraining scaling laws with pass@$k$ modeling used for test-time scaling, then jointly optimize pretraining and test-time decisions.
  arXiv  Detail & Related papers  (2026-04-01T21:17:32Z)
• DISCO: Diversifying Sample Condensation for Efficient Model Evaluation [59.01400190971061]
  Costly evaluation reduces inclusivity, slows the cycle of innovation, and worsens environmental impact.<n>We argue that promoting diversity among samples is not essential; what matters is to select samples thatmaximise diversity in model responses.<n>Our method, $textbfDiversifying Sample Condensation (DISCO)$, selects the top-k samples with the greatest model disagreements.
  arXiv  Detail & Related papers  (2025-10-09T08:53:59Z)
• Optimal Sparsity of Mixture-of-Experts Language Models for Reasoning Tasks [17.067788440109137]
  Mixture-of-Experts (MoE) models are now standard in state-of-the-art systems.<n>We investigate how MoE sparsity influences two distinct capability regimes: memorization skills and reasoning skills.
  arXiv  Detail & Related papers  (2025-08-26T04:31:28Z)
• Predictable Scale: Part I, Step Law -- Optimal Hyperparameter Scaling Law in Large Language Model Pretraining [59.369484219304866]
  We conduct an unprecedented empirical investigation training over 3,700 Large Language Models (LLMs) from scratch across 100 trillion tokens.<n>We establish a universal Scaling Law for hyperparameter optimization in LLM Pre-training, called Step Law.<n>Our estimated optima deviates from the global best performance found via exhaustive search by merely 0.094% on the test set.
  arXiv  Detail & Related papers  (2025-03-06T18:58:29Z)
• Scaling Laws in Linear Regression: Compute, Parameters, and Data [86.48154162485712]
  We study the theory of scaling laws in an infinite dimensional linear regression setup.<n>We show that the reducible part of the test error is $Theta(-(a-1) + N-(a-1)/a)$.<n>Our theory is consistent with the empirical neural scaling laws and verified by numerical simulation.
  arXiv  Detail & Related papers  (2024-06-12T17:53:29Z)
• Language models scale reliably with over-training and on downstream tasks [121.69867718185125]
  Scaling laws are useful guides for derisking expensive training runs. However, there remain gaps between current studies and how language models are trained. In contrast, scaling laws mostly predict loss on inference, but models are usually compared on downstream task performance.
  arXiv  Detail & Related papers  (2024-03-13T13:54:00Z)
• A Dynamical Model of Neural Scaling Laws [79.59705237659547]
  We analyze a random feature model trained with gradient descent as a solvable model of network training and generalization. Our theory shows how the gap between training and test loss can gradually build up over time due to repeated reuse of data.
  arXiv  Detail & Related papers  (2024-02-02T01:41:38Z)
• Winner-Take-All Column Row Sampling for Memory Efficient Adaptation of Language Model [89.8764435351222]
  We propose a new family of unbiased estimators called WTA-CRS, for matrix production with reduced variance. Our work provides both theoretical and experimental evidence that, in the context of tuning transformers, our proposed estimators exhibit lower variance compared to existing ones.
  arXiv  Detail & Related papers  (2023-05-24T15:52:08Z)
• SCA-Net: A Self-Correcting Two-Layer Autoencoder for Hyper-spectral Unmixing [3.918940900258555]
  We show that a two-layer autoencoder (SCA-Net) achieves error metrics that are scales apart ($10-5)$ from previously reported values. We also show that SCA-Net, based upon a bi-orthogonal representation, performs a self-correction when the the number of endmembers are over-specified.
  arXiv  Detail & Related papers  (2021-02-10T19:37:52Z)

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.