HypoAgent: An Agentic Framework for Interactive Abductive Hypothesis Generation over Knowledge Graphs

2026-05-29

AI Summary

This work addresses the limitations of existing hypothesis generation methods on knowledge graphs, which struggle to accurately capture user intent in multi-turn dialogues and lack fine-grained diagnostic capabilities for failed hypotheses. To overcome these challenges, the authors propose HypoAgent, a novel framework that introduces a multi-agent collaborative mechanism for interactive abductive reasoning. By orchestrating three specialized agents—intent recognition, hypothesis generation, and root cause analysis—the framework dynamically maps natural language intents into knowledge graph constraints, generates controllable hypotheses, and performs neighborhood-based attribution and repair upon failure. Integrating natural language understanding, graph querying, controllable generation, and structural probing, HypoAgent forms an end-to-end collaborative system that achieves state-of-the-art semantic similarity performance across single-turn, multi-turn, and unconditional settings on both commonsense and biomedical knowledge graphs.

Abstract

Abductive reasoning over knowledge graphs aims to generate logical hypotheses that explain observed entities or facts. Existing controllable hypothesis generation methods allow users to guide this process with explicit conditions, but they remain limited in interactive settings: they struggle to ground evolving natural-language intents across multi-turn dialogues and provide little fine-grained diagnosis when generated hypotheses fail. To address these limitations, we propose HypoAgent, an Agentic framework for interactive abductive Hypothesis Generation over knowledge graphs. HypoAgent integrates three agents: an Intent Recognition Agent that grounds user utterances and dialogue history into executable KG conditions, a Hypothesis Generation Agent that performs controllable hypothesis generation according to the extracted user intention, and a Root Cause Analysis Agent that diagnoses unreliable hypothesis fragments and leverages KG neighborhood probing to identify supported refinements. Experiments on commonsense and biomedical domain-specific knowledge graphs demonstrate that HypoAgent achieves state-of-the-art semantic similarity under single-turn, multi-turn, and unconditional settings. Our code is available at https://github.com/HKUST-KnowComp/HypoAgent.

Problem

• abductive reasoning
• knowledge graphs
• interactive hypothesis generation
• intent grounding
• root cause analysis

Innovation

• Agentic framework
• Abductive reasoning
• Knowledge graphs
• Interactive hypothesis generation
• Root cause analysis