Distill and Align Decomposition for Enhanced Claim Verification

TL;DR

This paper introduces a reinforcement learning method that jointly optimizes sentence decomposition and claim verification, significantly improving verification accuracy and subclaim quality across multiple settings.

Contribution

It presents a novel RL-based framework with multi-objective optimization for better decomposition and verification, outperforming existing prompt-based and RL methods.

Findings

Decomposer achieves 71.75% macro-F1 in claim verification.

Outperforms prompt-based approaches (+1.99, +6.24) and RL methods (+5.84).

Human evaluation confirms high subclaim quality.

Abstract

Complex claim verification requires decomposing sentences into verifiable subclaims, yet existing methods struggle to align decomposition quality with verification performance. We propose a reinforcement learning (RL) approach that jointly optimizes decomposition quality and verifier alignment using Group Relative Policy Optimization (GRPO). Our method integrates: (i) structured sequential reasoning; (ii) supervised finetuning on teacher-distilled exemplars; and (iii) a multi-objective reward balancing format compliance, verifier alignment, and decomposition quality. Across six evaluation settings, our trained 8B decomposer improves downstream verification performance to (71.75%) macro-F1, outperforming prompt-based approaches ((+1.99), (+6.24)) and existing RL methods ((+5.84)). Human evaluation confirms the high quality of the generated subclaims. Our framework enables smaller language models to achieve state-of-the-art claim verification by jointly optimising for verification accuracy and decomposition quality.