When Reasoning Goes Astray: Attention Dynamics of Uncontrolled Reasoning
cs.AI, cs.CL
Submitted: 2026-09-30
Updated: 2026-09-30
License: http://creativecommons.org/licenses/by/4.0/
The gist: Large reasoning models (LRMs) improve performance on complex tasks through extended reasoning, yet the same process can degenerate into redundant verification and persistent generation loops.
Terminology
Abstract
Large reasoning models (LRMs) improve performance on complex tasks through extended reasoning, yet the same process can degenerate into redundant verification and persistent generation loops. Such uncontrolled reasoning increases inference cost and creates risks of resource exhaustion and service degradation. However, existing mitigations largely truncate long outputs or react to surface repetition, and thus fail to distinguish normal thinking from uncontrolled reasoning or explain how benign reasoning degenerates into harmful behavior. In this paper, we operationalize LRM generation as four states and further introduce Reasoning-state Analysis via Dynamic Attention Responses (RADAR), which identifies the current reasoning state in real time and characterizes how effective reflection can develop into uncontrolled generation. Guided by RADAR's analysis, we further realign abnormal attention distributions toward patterns observed in normal requests and examine how this correction affects excessive reflection and persistent looping. Temporal analyses show that uncontrolled reasoning is characterized by attention distributions that deviate from normal generation, with abnormal trends becoming detectable before repetition begins. Correcting these deviations through Attention Realignment consistently reduces looping while largely preserving benign performance. Together, RADAR provide a mechanistic account of how reasoning becomes uncontrolled, offering actionable guidance for identifying critical failure stages and designing targeted runtime interventions.
Sources
- Training Verifiers to Solve Math Word Problems
- Missing Premise exacerbates Overthinking: Are Reasoning Models losing Critical Thinking Skill?
- Inverse Scaling in Test-Time Compute
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- Measuring Massive Multitask Language Understanding
- ThinkPrune: Pruning Long Chain-of-Thought of LLMs via Reinforcement Learning
- OverThink: Slowdown Attacks on Reasoning LLMs
- A Deep Reinforced Model for Abstractive Summarization
- GPQA: A Graduate-Level Google-Proof Q&A Benchmark
- Self-Consistency Improves Chain of Thought Reasoning in Language Models
- RECUR: Resource Exhaustion Attack via Recursive-Entropy Guided Counterfactual Utilization and Reflection
- Qwen3 Technical Report
- Resource Consumption Threats in Large Language Models
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Agents' Last Exam
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems