Mitigating Cognitive Inertia in Large Reasoning Models via Latent Spike Steering

TL;DR

This paper introduces STARS, a training-free method that detects and corrects cognitive inertia in large reasoning models by monitoring internal state dynamics and injecting cues, improving reasoning accuracy and efficiency.

Contribution

STARS is a novel, unsupervised framework that identifies reasoning transitions via latent state spikes and adaptively guides models without additional training.

Findings

Reduces redundant reasoning loops in LRMs.

Improves accuracy across multiple benchmarks.

Operates without extra fine-tuning or supervision.

Abstract

While Large Reasoning Models (LRMs) have achieved remarkable performance by scaling test-time compute, they frequently suffer from Cognitive Inertia, a failure pattern manifesting as either overthinking (inertia of motion) or reasoning rigidity (inertia of direction). Existing detection methods, typically relying on superficial textual heuristics like self-correction tokens, often fail to capture the model's unvoiced internal conflicts. To address this, we propose STARS (Spike-Triggered Adaptive Reasoning Steering), a training-free framework designed to rectify cognitive inertia by monitoring latent dynamics. STARS identifies Cognitive Pivots-critical moments of reasoning transition-by detecting distinct L2 distance spikes in the hidden states. Upon detection, the framework employs geometric trajectory analysis to diagnose the structural nature of the transition and injects state-aware language cues to steer the model in real-time. Our experiments across diverse benchmarks confirm that STARS efficiently curtails redundant loops while improving accuracy through the adaptive correction of erroneous trajectories. STARS offers a robust, unsupervised mechanism to optimize the reasoning process of LRMs without requiring additional fine-tuning.