Escaping Redundant Reasoning: Structure-Aware Search for Inference-Time LLMs

arXiv:2609.00738v1 Announce Type: new Abstract: Inference-time search with large language models (LLMs) often concentrates on a small set of structurally or semantically similar trajectories, leaving alternatives underexplored---a failure mode we call extit{reasoning basin collapse}. We introduce BASIN, a training-free, structure-aware selection method that groups reasoning states into basins and penalizes repeated visits to the same strategy, thereby reallocating search across genuinely distinct reasoning paths under a fixed compute budget. Under matched inference budgets, BASIN improves over Tree of Thoughts (ToT) by up to +22pp on Game of 24 and +6.7pp on MuSR. A quality-aware variant, QA-BASIN, further improves robustness by preserving high-quality basins when unconditional diversification over-explores. To explain when basin-aware selection helps, we introduce the redundancy gap Delta, which measures how differently search concentrates for correct versus incorrect predictions: standard ToT often operates near Delta approx 0, while BASIN consistently shifts Delta positive. More broadly, BASIN suggests structure-aware selection as a simple and general approach to improving inference-time reasoning. Code can be found at https://github.com/GitHubLuCheng/basin.

Related

• Dual-Dimensional Consistency: Balancing Budget and Quality in Adaptive Inference-Time Scaling
• Scaling-Aware Adapter for Structure-Grounded LLM Reasoning
• DecodeShare: Tracing the Shared Subspace of LLM Decode-Time Decisions
• AutoCRAT: Within-trajectory Joint Control of Stochasticity and Compute for LLM Reasoning
• Extracting Search Trees from LLM Reasoning Traces Reveals Myopic Planning

Source: arXiv cs.AI | 2026-09-02