CEDAR-GRPO: Process-Aware Reinforcement Learning for General Abductive Reasoning in LLMs

2026-08-14

AI Summary

This study addresses the challenge of cross-task transfer in abductive reasoning for large language models by proposing a process-aware reinforcement learning framework. The approach integrates answer correctness with an abductive reward that balances evidence coverage and directional alignment, leveraging Group Relative Policy Optimization (GRPO) alongside multi-task mixed training during post-training to acquire generalizable abductive capabilities. Experimental results demonstrate that the model significantly outperforms baselines across all unseen tasks, achieving an average improvement of 7.4 points and a maximum gain of 30.8 points. These findings validate the framework’s effectiveness in enhancing generalization and transferability, establishing a novel paradigm for complex reasoning tasks.

Abstract

Abductive reasoning, often characterized as inference to the best explanation, is central to explanation under uncertainty, from everyday sense-making and investigation to scientific discovery. Yet LLM research has mostly studied abduction through narrow, task-specific benchmarks, making it unclear whether observed gains transfer beyond the benchmark family used for training or evaluation. We ask whether RL post-training can improve abduction as a transferable reasoning capability. We introduce CEDAR-GRPO, a process-aware framework that combines final-answer correctness with abductive rewards for evidence coverage and evidence-to-explanation directionality. Four open-weight LLMs are post-trained on a controlled, domain-neutral mixture of abductive hypothesis-generation and hypothesis-selection tasks. We evaluate them on 11 unseen tasks spanning hypothesis selection, missing-fact generation, defeasible inference, long-context investigation, clinical reasoning, code debugging, and non-abductive controls. CEDAR- GRPO improves every model on every held-out task over both base models and correctness-only GRPO, with average gains of 7.4 and 2.7 points, respectively, and a maximum gain of 30.8 points. Ablations confirm that RL, abductive reward design, and task diversity each contribute to transfer. Process-level metrics further show stronger abductive behavior, including exploration of alternatives, elimination of rivals, backtracking, and uncertainty marking.

Problem

• Abductive Reasoning
• Transferable Reasoning Capability
• Reinforcement Learning Post-training
• Large Language Models

Innovation

• Process-Aware Reinforcement Learning
• Abductive Reasoning
• CEDAR-GRPO
• Transferable Reasoning
• Abductive Rewards