Tab-TRM: Tiny Recursive Model for Insurance Pricing on Tabular Data

Paper: arXiv 2601.07675

Authors: Kishan Padayachy, Ronald Richman, Mario V. Wüthrich

Abstract

We introduce Tab-TRM (Tabular-Tiny Recursive Model), a network architecture that adapts the recursive latent reasoning paradigm of Tiny Recursive Models (TRMs) to insurance modeling. Drawing inspiration from both the Hierarchical Reasoning Model (HRM) and its simplified successor TRM, the Tab-TRM model makes predictions by reasoning over the input features. It maintains two learnable latent tokens - an answer token and a reasoning state - that are iteratively refined by a compact, parameter-efficient recursive network. The recursive processing layer repeatedly updates the reasoning state given the full token sequence and then refines the answer token, in close analogy with iterative insurance pricing schemes. Conceptually, Tab-TRM bridges classical actuarial workflows - iterative generalized linear model fitting and minimum-bias calibration - on the one hand, and modern machine learning, in terms of Gradient Boosting Machines, on the other.

Complexity vs Empirical Score

• Math Complexity: 5.5/10
• Empirical Rigor: 7.5/10
• Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: This paper presents an innovative adaptation of Tiny Recursive Models (TRM) to tabular insurance pricing, bridging classical actuarial iterative algorithms with modern latent reasoning architectures. The methodology is clearly explained, mathematically sound, and empirically validated against the benchmark French MTPL dataset with strong performance.

Research Flowchart

  flowchart TD
    A[Research Goal: Adapt Recursive Latent Reasoning for Insurance Pricing on Tabular Data] --> B[Data Inputs: Tabular Features & Insurance Liabilities Data]
    B --> C[Tokenization: Initialize Answer Token & Reasoning State]
    C --> D[Computational Core: Iterative Refinement via Compact Recursive Network]
    D --> E[Conceptual Bridge: Connect GLM Calibration, Minimum-Bias, & GBMs]
    E --> F[Key Outcome: Tab-TRM Architecture for Parameter-Efficient Actuarial Pricing]