Self-Verification Dilemma: Experience-Driven Suppression of Overused Checking in LLM Reasoning
This addresses inefficiency in LLM reasoning for users by reducing computational costs, but it is incremental as it optimizes an existing method rather than introducing a new paradigm.
The paper tackles the problem of excessive self-verification steps in Large Reasoning Models, which are often confirmatory rather than corrective, and proposes an experience-driven framework that reduces token usage by up to 20.3% while maintaining or improving accuracy.
Large Reasoning Models (LRMs) achieve strong performance by generating long reasoning traces with reflection. Through a large-scale empirical analysis, we find that a substantial fraction of reflective steps consist of self-verification (recheck) that repeatedly confirm intermediate results. These rechecks occur frequently across models and benchmarks, yet the vast majority are confirmatory rather than corrective, rarely identifying errors and altering reasoning outcomes. This reveals a mismatch between how often self-verification is activated and how often it is actually useful. Motivated by this, we propose a novel, experience-driven test-time framework that reduces the overused verification. Our method detects the activation of recheck behavior, consults an offline experience pool of past verification outcomes, and estimates whether a recheck is likely unnecessary via efficient retrieval. When historical experience suggests unnecessary, a suppression signal redirects the model to proceed. Across multiple model and benchmarks, our approach reduces token usage up to 20.3% while maintaining the accuracy, and in some datasets even yields accuracy improvements.