Structured grounding improves reasoning efficiency in language models
Better Reasoning, Not Just Bigger Models: A Structure-Grounded Framework for Efficient and Reliable Inference
Artificial IntelligenceInformation Retrieval
Summary
Language models have become better at generating fluent text, but they sometimes struggle to reason correctly, especially when information is incomplete. The authors propose a method called Structure-Grounded Uncertainty Reduction (SGUR), which improves reasoning by representing knowledge as graphs and actively reducing uncertainty by focusing on the most important information. They show that this approach helps smaller models perform as well as much larger models on a medical diagnosis task, using fewer tokens and providing more reliable reasoning.
What this means in practice
- •For medical ai developers: Improve diagnostic systems by using structure-based reasoning to reduce information needed and boost accuracy in sequential medical diagnosis.
- •For chatbot designers: Build conversational agents that can reason more reliably under uncertainty by actively selecting critical information to query.
Authors
Amirhossein Aghaei, Liangyan Li, Shiyi Yang, Shaohua Wang, Yan Wang
Abstract
Scaling large language models has substantially improved linguistic fluency, yet reliable autonomous agents require more than fluent generation. In multi-turn inference, current agentic systems often produce semantically plausible but structurally ungrounded reasoning, especially under incomplete observations. We propose Structure-Grounded Uncertainty Reduction (SGUR), a training-free framework that complements token-level generation with active uncertainty reduction over structured hypothesis spaces. SGUR represents incomplete observations as a query graph and aligns them with candidate hypothesis graphs through a prior-weighted partial graph matching mechanism, where database-derived empirical priors assign greater importance to decision-relevant relations rather than treating all graph edges uniformly. The resulting inference state is iteratively refined by an active information-gain strategy that selects the next most informative node to query, systematically pruning the candidate space. We instantiate SGUR on a challenging PrimeKG-derived sequential medical diagnosis benchmark, demonstrating that structural grounding can partially substitute for model scale: starting from only two observed symptoms, SGUR achieves 92.5% Hit@3, outperforming the strongest Graph-RAG baseline by 9.4 percentage points while using approximately 5x fewer tokens. Notably, SGUR enables smaller backbone models to match or exceed the performance of models up to 10x larger, suggesting that structural grounding provides a more robust and efficient paradigm for scaling agent performance than relying exclusively on increasing model size.