AdaR: A Framework for Equipping LLMs with Adaptive Reasoning

ACL ARR 2026 January Submission9479 Authors

Keywords: Synthetic Data, Chain-of-Thought, Mathematical Reasoning

Abstract:

Mathematical reasoning is a primary indicator of large language models (LLMs) intelligence.
However, existing LLMs exhibit failures of robustness and generalization.
This paper attributes these deficiencies to spurious reasoning, wherein generated reasoning traces bear negligible causal connection to answers.
To address this challenge, we propose the AdaR framework to equip LLMs with adaptive reasoning, wherein models rely on problem-solving logic to produce answers.
AdaR automatically synthesizes logically equivalent queries by varying variable values, and trains models with Reinforcement Learning with Verifiable Rewards (RLVR) on these data to penalize spurious logic while encouraging adaptive logic.
To ensure data quality, we extract the problem-solving logic from the original query and generate the corresponding answer by code execution and then apply sanity check.
Experimental results demonstrate that AdaR achieves substantial improvement in mathematical reasoning while maintaining high data efficiency.
Furthermore, even for advanced LLMs, there still exists robustness and generalization deficiencies, which our work effectively mitigates.

Paper Type: Long

Research Area: Mathematical, Symbolic, Neurosymbolic, and Logical Reasoning

Research Area Keywords: Efficient/Low-Resource Methods for NLP, NLP Applications

Contribution Types: Approaches to low-resource settings, Data resources

Languages Studied: English

Submission Number: 9479