Is Your Model Thinking or Just Stagnating? PUMA: Diagnosing Reasoning Pathology via Phase-Momentum Alignment

• URL: http://arxiv.org/abs/2607.17188v1
• Date: Sun, 19 Jul 2026 11:00:59 GMT
• Title: Is Your Model Thinking or Just Stagnating? PUMA: Diagnosing Reasoning Pathology via Phase-Momentum Alignment
• Abstract summary: Test-time scaling empowers Large Reasoning Models (LRMs) to tackle complex tasks via extensive Chain-of-Thought (CoT)<n>This paper introduces PUMA, a training-free framework employing a tiered diagnostic architecture.<n>Experiments on LRMs spanning 1.5B to 32B demonstrate that PUMA consistently outperforms state-of-the-art baselines across diverse benchmarks.
• Score: 13.9489759632504
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Test-time scaling empowers Large Reasoning Models (LRMs) to tackle complex tasks via extensive Chain-of-Thought (CoT). However, this often induces the "overthinking" paradox, where redundant reasoning increases computational overhead without guaranteeing accuracy. Existing test-time efficiency optimization methods primarily fall into two categories: information-theoretic approaches, which are prone to "deceptive convergence" where low uncertainty masks hallucinations, and latent representation analyses, which are often post-hoc, lacking the real-time sensitivity for dynamic reasoning. To bridge this gap, we first posit the Phase-Momentum Alignment Hypothesis, asserting that reasoning correctness hinges on the temporal synchronization between geometric momentum and uncertainty resolution. We then theoretically formulate the Cognitive-Energy Model to characterize these dynamics through two orthogonal dimensions: Geometric Cognitive Effort, quantified by latent velocity and tortuosity, and Entropic Cognitive Uncertainty. To operationalize this, we introduce PUMA (Phase-Uncertainty Momentum Alignment), a training-free framework employing a tiered diagnostic architecture. By coupling lightweight phase monitoring with event-triggered geometric analysis, PUMA effectively distinguishes active exploration from passive stagnation, enabling precise interventions through adaptive truncation or corrective measures. Extensive experiments on LRMs spanning 1.5B to 32B demonstrate that PUMA consistently outperforms state-of-the-art baselines across diverse benchmarks, achieving a superior accuracy-efficiency trade-off and robust cross-domain generalization.

Related papers

• Entropy-Gradient Inversion: Moving Toward Internal Mechanism of Large Reasoning Models [47.195670444638715]
  We identify and formally define textbfEntropy-Gradient Inversion, a robust correlation between token entropy and logit gradients.<n>We propose textbfCorrelation-Regularized Group Policy Optimization (CorR-PO), which embeds this inversion signature into reinforcement learning.<n>Experiments on various reasoning benchmarks across multiple model scales show CorR-PO consistently outperforms state-of-the-art baselines.
  arXiv  Detail & Related papers  (2026-05-18T02:41:53Z)
• Cognitive Friction: A Decision-Theoretic Framework for Bounded Deliberation in Tool-Using Agents [0.0]
  Unconstrained tool-using agents must decide which information source to query and act on.<n>We propose the Triadic Cognitive Architecture (TCA), a decision-theoretic framework that formalizes these failure modes via cognitive friction.<n>We validate TCA in two controlled environments designed to isolate stopping quality, action selection under congestion, and temporal urgency.
  arXiv  Detail & Related papers  (2026-03-31T17:30:25Z)
• Dual-Gated Epistemic Time-Dilation: Autonomous Compute Modulation in Asynchronous MARL [0.0]
  Multi-Agent Reinforcement Learning (MARL) algorithms achieve unprecedented successes across complex continuous domains.<n>Their standard deployment strictly adheres to a synchronous operational paradigm.<n>This dense throughput acts as a fundamental barrier to physical deployment on edge-devices where thermal and metabolic budgets are highly constrained.<n>We propose Epistemic Time-Dilation MAPPO (ETD-MAPPO), augmented with a Dual-Gated Epistemic Trigger.
  arXiv  Detail & Related papers  (2026-03-24T21:19:06Z)
• Anatomy of a Lie: A Multi-Stage Diagnostic Framework for Tracing Hallucinations in Vision-Language Models [62.932580559941414]
  Vision-Language Models (VLMs) frequently "hallucinate" - generate plausible yet factually incorrect statements.<n>We propose a new paradigm for diagnosing hallucinations, recasting them from static output errors into dynamic pathologies of a model's computational cognition.
  arXiv  Detail & Related papers  (2026-03-16T17:20:38Z)
• Mitigating Cognitive Inertia in Large Reasoning Models via Latent Spike Steering [12.332146893333949]
  Large Reasoning Models (LRMs) have achieved remarkable performance by scaling test-time compute.<n>LRMs frequently suffer from Cognitive Inertia, a failure pattern manifesting as either overthinking (inertia of motion) or rigidity (inertia of direction)
  arXiv  Detail & Related papers  (2026-01-30T02:47:12Z)
• Agentic Uncertainty Quantification [76.94013626702183]
  We propose a unified Dual-Process Agentic UQ (AUQ) framework that transforms verbalized uncertainty into active, bi-directional control signals.<n>Our architecture comprises two complementary mechanisms: System 1 (Uncertainty-Aware Memory, UAM), which implicitly propagates verbalized confidence and semantic explanations to prevent blind decision-making; and System 2 (Uncertainty-Aware Reflection, UAR), which utilizes these explanations as rational cues to trigger targeted inference-time resolution only when necessary.
  arXiv  Detail & Related papers  (2026-01-22T07:16:26Z)
• CoG: Controllable Graph Reasoning via Relational Blueprints and Failure-Aware Refinement over Knowledge Graphs [53.199517625701475]
  CoG is a training-free framework inspired by Dual-Process Theory that mimics the interplay between intuition and deliberation.<n>CoG significantly outperforms state-of-the-art approaches in both accuracy and efficiency.
  arXiv  Detail & Related papers  (2026-01-16T07:27:40Z)
• A Generalized Adaptive Joint Learning Framework for High-Dimensional Time-Varying Models [0.8594140167290097]
  This article introduces Adaptive Joint Learning (AJL), a regularization framework designed to simultaneously perform functional variable selection and structural changepoint detection.<n>The analysis uncovers synchronized phase transitions in disease progression and identifies a parsimonious set of time-varying prognostic markers.
  arXiv  Detail & Related papers  (2026-01-08T02:07:49Z)
• Efficient Thought Space Exploration through Strategic Intervention [54.35208611253168]
  We propose a novel Hint-Practice Reasoning (HPR) framework that operationalizes this insight through two synergistic components.<n>The framework's core innovation lies in Distributional Inconsistency Reduction (DIR), which dynamically identifies intervention points.<n> Experiments across arithmetic and commonsense reasoning benchmarks demonstrate HPR's state-of-the-art efficiency-accuracy tradeoffs.
  arXiv  Detail & Related papers  (2025-11-13T07:26:01Z)
• Emergence of Superposition: Unveiling the Training Dynamics of Chain of Continuous Thought [64.43689151961054]
  We theoretically analyze the training dynamics of a simplified two-layer transformer on the directed graph reachability problem.<n>Our analysis reveals that during training using continuous thought, the index-matching logit will first increase and then remain bounded under mild assumptions.
  arXiv  Detail & Related papers  (2025-09-27T15:23:46Z)
• Adaptive Termination for Multi-round Parallel Reasoning: An Universal Semantic Entropy-Guided Framework [12.361554676966552]
  Recent advances in large language models (LLMs) have accelerated progress toward artificial general intelligence.<n>We aim to design a flexible test-time collaborative inference framework that exploits the complementary strengths of both sequential and parallel reasoning paradigms.
  arXiv  Detail & Related papers  (2025-07-09T13:28:35Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.