Paper 2510.06217
TaTToo: Tool-Grounded Thinking PRM for Test-Time Scaling in Tabular Reasoning
- Published
- Oct 2025
- Research lab
- Independent
- Citations
- 10
- GitHub
- Not linked
01 In brief
Summary
TATTOO is a novel table-grounded Process Reward Model (PRM) designed to improve test-time scaling (TTS) for large reasoning models (LRMs) on tabular reasoning tasks.
The authors identify that existing PRMs fail on table-specific operations like sub-table retrieval and schema interaction, leading to performance bottlenecks.
TATTOO addresses this by integrating tool-based verification and providing step-level rewards that distinguish between table operations and inner reasoning.
A scalable data curation pipeline generates over 60k high-quality step-level annotations using table verification rationales and tool executions.
TATTOO is trained via a dual-stage paradigm: supervised fine-tuning (SFT) followed by reinforcement learning (RL) with tool-grounded reward shaping.
Experiments on five benchmarks (TableBench, WTQ, MMQA) show TATTOO improves downstream policy LRMs by 30.9% at inference, outperforming stronger PRMs like Qwen-2.5-Math-PRM-72B with only 8B parameters.
It also generalizes across TTS strategies like Beam Search and DVTS.
Ablations confirm the RL stage adds a 10.2% improvement over SFT alone, and the tool-grounding reward component is critical.
02 From the paper
Abstract
Process Reward Models (PRMs) have recently emerged as a powerful framework for enhancing the reasoning capabilities of large reasoning models (LRMs), particularly in the context of test-time scaling (TTS). However, their potential for supervising LRMs on tabular reasoning domains remains underexplored. Through detailed empirical analyses, we identify that existing PRMs, though widely adopted for supervising text-only reasoning steps, struggle with table-specific operations such as sub-table retrieval and schema interaction, leading to critical performance bottlenecks. To address this limitation, we propose TaTToo, a novel table-grounded PRM framework that (i) reasons explicitly over tabular reasoning steps and (ii) integrates tool-based verification to provide precise reward supervision. Concretely, we first design a scalable data curation pipeline that constructs over 60k high-quality step-level annotations by integrating table verification rationales with tool-based executions. Building on the collected data, we train TaTToo with a dual-stage paradigm: cold-start supervised fine-tuning to capture tool-use reasoning patterns, followed by reinforcement learning with tool-grounded reward shaping to align our model with table-based verification. We provide a comprehensive evaluation of the policy improvement induced by our newly designed PRM. Across 5 challenging tabular reasoning benchmarks covering numerical reasoning, fact-checking, and data analysis, TaTToo improves downstream policy LRMs by 30.9% at inference, surpasses strong PRM baselines such as Qwen-2.5-Math-PRM-72B with only 8B parameters, and demonstrates strong generalizability across diverse TTS strategies.