Internal Flow Signatures for Self-Checking and Refinement in LLMs

• URL: http://arxiv.org/abs/2602.01897v1
• Date: Mon, 02 Feb 2026 10:05:54 GMT
• Title: Internal Flow Signatures for Self-Checking and Refinement in LLMs
• Abstract summary: We introduce emphinternal flow signatures that audit decision formation from depthwise dynamics at a fixed inter-block monitoring boundary.<n>The method stabilizes token-wise motion via bias-centered monitoring, then summarizes trajectories in compact emphmoving readout-aligned subspaces.<n>A lightweight GRU validator trained on these signatures performs self-checking without modifying the base model.
• Score: 8.540143367732949
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large language models can generate fluent answers that are unfaithful to the provided context, while many safeguards rely on external verification or a separate judge after generation. We introduce \emph{internal flow signatures} that audit decision formation from depthwise dynamics at a fixed inter-block monitoring boundary. The method stabilizes token-wise motion via bias-centered monitoring, then summarizes trajectories in compact \emph{moving} readout-aligned subspaces constructed from the top token and its close competitors within each depth window. Neighboring window frames are aligned by an orthogonal transport, yielding depth-comparable transported step lengths, turning angles, and subspace drift summaries that are invariant to within-window basis choices. A lightweight GRU validator trained on these signatures performs self-checking without modifying the base model. Beyond detection, the validator localizes a culprit depth event and enables a targeted refinement: the model rolls back to the culprit token and clamps an abnormal transported step at the identified block while preserving the orthogonal residual. The resulting pipeline provides actionable localization and low-overhead self-checking from internal decision dynamics. \emph{Code is available at} \texttt{github.com/EavnJeong/Internal-Flow-Signatures-for-Self-Checking-and-Refinement-in-LLMs}.

Related papers

• Probabilistic Reachable-Action Verification of Visuomotor Policies via Set-Based Training [7.193186752397934]
  Large visual encoders make end-to-end set propagation computationally expensive and excessively conservative.<n>We freeze the visual encoder and confine set propagation to a low-dimensional interface between it and the downstream policy.<n>Propagating this set through the policy with zonotopes yields a terminal output-enclosure width that set-based training optimize directly.
  arXiv  Detail & Related papers  (2026-08-03T17:32:08Z)
• From Open Loop to Closed Loop: A Test-Time Iterative Optimization Framework for Reference-Consistent Image Generation [108.02425846269234]
  controllable image generation has made significant strides by incorporating visual reference conditions.<n>Existing methods predominantly operate as open-loop systems.<n>We propose a novel test-time iterative optimization framework that reformulates reference-consistent generation as a closed-loop dynamic tracking problem.
  arXiv  Detail & Related papers  (2026-07-06T05:36:30Z)
• CIAware-Bench: Benchmarking Control Intervention Awareness Across Frontier LLMs [100.38986535324284]
  We introduce textbfCIAware-Bench, a benchmark for measuring textbfcontrol textbfintervention (CI) awareness across frontier models.<n>CIAware-Bench tests whether models can distinguish their own trajectories from those modified by a control intervention.
  arXiv  Detail & Related papers  (2026-06-09T16:24:16Z)
• StainFlow: Entity-Stain Tracking and Evidence Linking for Process Rewards in GUI Agents [67.03593791535786]
  Reinforcement Learning (RL) has become a promising approach for improving GUI Agents in long-horizon, digital environments.<n>To mitigate this issue, recent studies introduce Process Reward Models (PRMs)<n>PRMs provide finer-grained training feedback through global milestone verification or local step-level evaluation.<n>We propose StainFlow, an entity-stain-flow process reward model for GUI Agents.
  arXiv  Detail & Related papers  (2026-06-05T08:17:28Z)
• Spectral Guidance for Flexible and Efficient Control of Diffusion Models [25.912824880444262]
  We introduce Spectral Guidance, a framework for controlling diffusion models by leveraging the intrinsic geometry of the generative process.<n>We characterize them as the singular functions of a conditional expectation operator and show that they can be learned via a self-supervised objective.
  arXiv  Detail & Related papers  (2026-05-27T15:11:09Z)
• Semantic Robustness Probing via Inpainting: An Interactive Tool for Safety-Critical Object Detection [0.47248250311484113]
  Testing object detectors in safety-critical domains requires semantically meaningful probes beyond pixel-level corruptions.<n>We present SemProbe, a tool for semantic robustness probing.
  arXiv  Detail & Related papers  (2026-05-26T15:15:38Z)
• The Path Matters: Learning a Token-Commitment Policy for Diffusion Language Models [52.93186090124315]
  We argue that token commitment can be learned as a reusable trace-state policy.<n>We introduce TraceLock, a lightweight plug-in controller that instantiates this policy for a frozen diffusion language model.
  arXiv  Detail & Related papers  (2026-05-23T18:23:46Z)
• MoASE++: Mixture of Activation Sparsity Experts with Domain-Adaptive On-policy Distillation for Continual Test Time Adaptation [65.068801413044]
  Continual test-time adaptation adapts a source-pretrained model to non-stationary, unlabeled target streams.<n>We introduce MoASE, a mixture-of-experts that disentangles domain-agnostic structure from domain-specific texture.
  arXiv  Detail & Related papers  (2026-05-18T01:52:12Z)
• OmniVL-Guard Pro: A Tool-Augmented Agent for Omnibus Vision-Language Forensics [63.13200245209719]
  We propose textbf OmniVL-Guard Pro, a tool-augmented agent that extends unified forensics from closed-world prediction to open-world clues-driven reasoning.<n>To generate high-quality tool-reasoning trajectories, we introduce textbfTree-Structured Self-Evolving Tool Trajectory Generation.<n>We also propose textbfChecker-Guided Agentic Reinforcement Learning, which provides process-level supervision to penalize cases where the answer is correct but the reasoning is distorted.
  arXiv  Detail & Related papers  (2026-05-16T12:26:04Z)
• Sparse Visual Thought Circuits in Vision-Language Models [2.5754366051855837]
  We develop a causal pipeline to localize and test sparse visual thought circuits in Qwen3-VL-8B.<n>We train SAEs at this layer, construct task-selective sets via an explicit rule, and perform inference time scaling and ablation while quantifying accuracy and drift.
  arXiv  Detail & Related papers  (2026-03-26T06:24:36Z)
• Beyond Input Guardrails: Reconstructing Cross-Agent Semantic Flows for Execution-Aware Attack Detection [32.301679396929536]
  We propose SysName, a framework that shifts the defensive paradigm from static input filtering to execution-aware analysis.<n>SysName synthesizes fragmented operational primitives into contiguous behavioral trajectories, enabling a holistic view of system activity.<n> Empirical evaluations demonstrate that SysName effectively detects over ten distinct compound attack vectors, achieving F1-scores of 85.3% and 66.7% for node-level and path-level end-to-end attack detection, respectively.
  arXiv  Detail & Related papers  (2026-03-04T01:59:16Z)
• IoUCert: Robustness Verification for Anchor-based Object Detectors [58.35703549470485]
  We introduce IoUCert, a novel formal verification framework designed specifically to overcome these bottlenecks in anchor-based object detection architectures.<n>We show that our method enables the robustness verification of realistic, anchor-based models including SSD, YOLOv2, and YOLOv3 variants against various input perturbations.
  arXiv  Detail & Related papers  (2026-03-03T14:36:46Z)
• Rubrics as an Attack Surface: Stealthy Preference Drift in LLM Judges [39.6038287387348]
  LLM-based judges' behavior is guided by natural rubrics and validated on benchmarks.<n>We identify a previously underrecognized vulnerability in this workflow, which we term-Induced Preference Drift (RIPD)<n>We show this vulnerability can be exploited through preference attacks, in which benchmark-compliant edits steer judgments away from a fixed human or trusted reference on target domains.
  arXiv  Detail & Related papers  (2026-02-14T03:19:14Z)
• When Domain Pretraining Interferes with Instruction Alignment: An Empirical Study of Adapter Merging in Medical LLMs [0.6345523830122167]
  Large language models can exhibit surprising adapter interference when combining domain adaptation and instruction alignment.<n>We study a two-stage LoRA pipeline for medical LLMs, where domain-oriented pre-training (PT) and supervised fine-tuning (SFT) are trained separately and later merged.
  arXiv  Detail & Related papers  (2026-01-26T10:54:06Z)
• Reasoning Stabilization Point: A Training-Time Signal for Stable Evidence and Shortcut Reliance [0.0]
  We define explanation drift as the epoch-to-epoch change in normalized token attributions on a fixed probe set.<n>RSP is computed from within-run drift dynamics and requires no tuning on out-of-distribution data.
  arXiv  Detail & Related papers  (2026-01-12T17:48:05Z)
• TIE: A Training-Inversion-Exclusion Framework for Visually Interpretable and Uncertainty-Guided Out-of-Distribution Detection [11.599035626374409]
  Deep neural networks often struggle to recognize when an input lies outside their training experience, leading to unreliable and overconfident predictions.<n>We propose textbfTIE: a Training--Inversion--Exclusion framework for visually interpretable and uncertainty-guided anomaly detection.
  arXiv  Detail & Related papers  (2025-11-28T22:06:01Z)
• BlockCert: Certified Blockwise Extraction of Transformer Mechanisms [0.0]
  We introduce BlockCert, a framework for certified blockwise extraction of transformer mechanisms.<n>We formalize a simple Lipschitz-based composition theorem in Lean 4 that lifts these local guarantees to a global deviation bound.<n>Our results suggest that blockwise extraction with explicit certificates is feasible for real transformer language models.
  arXiv  Detail & Related papers  (2025-11-20T06:04:34Z)
• PIXEL: Adaptive Steering Via Position-wise Injection with eXact Estimated Levels under Subspace Calibration [17.225716209866086]
  We propose a position-wise activation steering framework for large language models (LLMs) on the web.<n>PIXEL learns a property-aligned subspace from dual views and selects intervention strength via a constrained geometric objective.<n>PIXEL consistently improves attribute alignment while preserving model general capabilities.
  arXiv  Detail & Related papers  (2025-10-11T13:13:34Z)
• DiffuSpec: Unlocking Diffusion Language Models for Speculative Decoding [66.40658898418316]
  We present DiffuSpec, a training-free drop-in framework that uses a pretrained diffusion language model (DLM) to produce multi-token drafts in a single forward pass.<n>Across benchmarks, DiffuSpec yields up to 3x wall-clock speedup, establishing diffusion-based drafting as a robust alternative to autoregressive drafters for speculative decoding.
  arXiv  Detail & Related papers  (2025-09-28T07:00:15Z)
• Unsupervised Conformal Inference: Bootstrapping and Alignment to Control LLM Uncertainty [49.19257648205146]
  We propose an unsupervised conformal inference framework for generation.<n>Our gates achieve close-to-nominal coverage and provide tighter, more stable thresholds than split UCP.<n>The result is a label-free, API-compatible gate for test-time filtering.
  arXiv  Detail & Related papers  (2025-09-26T23:40:47Z)
• GrAInS: Gradient-based Attribution for Inference-Time Steering of LLMs and VLMs [56.93583799109029]
  GrAInS is an inference-time steering approach that operates across both language-only and vision-language models and tasks.<n>During inference, GrAInS hidden activations at transformer layers guided by token-level attribution signals, and normalizes activations to preserve representational scale.<n>It consistently outperforms both fine-tuning and existing steering baselines.
  arXiv  Detail & Related papers  (2025-07-24T02:34:13Z)
• LoRA-BAM: Input Filtering for Fine-tuned LLMs via Boxed Abstraction Monitors over LoRA Layers [3.674863913115432]
  Fine-tuning large language models (LLMs) improves performance on domain-specific tasks but can lead to overfitting, making them unreliable on out-of-distribution (OoD) queries.<n>We propose LoRA-BAM - a method that adds OoD detection monitors to the LoRA layer using boxed abstraction to filter questions beyond the model's competence.
  arXiv  Detail & Related papers  (2025-06-01T12:58:32Z)
• Lazy Layers to Make Fine-Tuned Diffusion Models More Traceable [70.77600345240867]
  A novel arbitrary-in-arbitrary-out (AIAO) strategy makes watermarks resilient to fine-tuning-based removal. Unlike the existing methods of designing a backdoor for the input/output space of diffusion models, in our method, we propose to embed the backdoor into the feature space of sampled subpaths. Our empirical studies on the MS-COCO, AFHQ, LSUN, CUB-200, and DreamBooth datasets confirm the robustness of AIAO.
  arXiv  Detail & Related papers  (2024-05-01T12:03:39Z)
• Self-Supervised 3D Scene Flow Estimation and Motion Prediction using Local Rigidity Prior [100.98123802027847]
  We investigate self-supervised 3D scene flow estimation and class-agnostic motion prediction on point clouds. We generate pseudo scene flow labels for self-supervised learning through piecewise rigid motion estimation. Our method achieves new state-of-the-art performance in self-supervised scene flow learning.
  arXiv  Detail & Related papers  (2023-10-17T14:06:55Z)
• Semi-DETR: Semi-Supervised Object Detection with Detection Transformers [105.45018934087076]
  We analyze the DETR-based framework on semi-supervised object detection (SSOD) We present Semi-DETR, the first transformer-based end-to-end semi-supervised object detector. Our method outperforms all state-of-the-art methods by clear margins.
  arXiv  Detail & Related papers  (2023-07-16T16:32:14Z)

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.