Patch-Effect Graph Kernels for LLM Interpretability

• URL: http://arxiv.org/abs/2605.06480v1
• Date: Thu, 07 May 2026 16:03:47 GMT
• Title: Patch-Effect Graph Kernels for LLM Interpretability
• Abstract summary: Mechanistic interpretability aims to reverse-engineer transformer computations by identifying causal circuits through activation patching.<n>We propose a framework that reframes mechanistic analysis as a graph machine-learning problem by representing activation-patching profiles as patch-effect graphs over model components.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Mechanistic interpretability aims to reverse-engineer transformer computations by identifying causal circuits through activation patching. However, scaling these interventions across diverse prompts and task families produces high-dimensional, unstructured datasets that are difficult to compare systematically. We propose a framework that reframes mechanistic analysis as a graph machine-learning problem by representing activation-patching profiles as patch-effect graphs over model components. We introduce three graph-construction methods: direct-influence via causal mediation, partial-correlation, and co-influence and apply graph kernels to analyze the resulting structures. Evaluating this approach on GPT-2 Small using Indirect Object Identification (IOI) and related tasks, we find that patch-effect graphs preserve discriminative structural signals. Specifically, localized edge-slot features provide higher classification accuracy than global graph-shape descriptors. A screened paired-patching validation suggests that CI and PC selected candidate edges correspond to stronger activation-influence effects than random or low-rank candidates. Crucially, by evaluating these representations against rigorous prompt-only and raw patch-effect controls, we make the evidential scope of the benchmark explicit: graph features compress structured patching signal, while raw tensors and surface cues define strong baselines that any circuit-level claim should address. Ultimately, our framework provides a compression and evaluation pipeline for comparing patching-derived structures under controlled baselines, separating robust slice-discriminative evidence from stronger task-general causal-circuit claims.

Related papers

• DeCoRAG: Cognitive Decoupling and Semantic-Aware Cropping for Complex Document Understanding [9.340832029984703]
  We introduce DeCoRAG, a multimodal Graph RAG pipeline that shifts knowledge processing from coupled visual-semantic reasoning to "Cognitive Decoupling"<n>DeCoRAG improves the semantic pass rate by up to 12.5 percentage points over the strongest baseline.<n>RAP-Crop reduces offline graph-construction prompt tokens by 40.8% without sacrificing end-to-end accuracy.
  arXiv  Detail & Related papers  (2026-07-27T15:28:02Z)
• VeriGraph: Towards Verifiable Data-Analytic Agents [71.17693022916896]
  LLM-based agents have demonstrated strong capabilities in data-intensive analytical tasks, yet their outputs are rarely verifiable.<n>We propose VeriGraph, a traceable neuro-symbolic reasoning framework that enables agents to construct an explicit heterogeneous evidence directed acyclic graph.<n>Our results suggest that explicit evidence-graph construction is a promising path toward verifiable data-analytic agents.
  arXiv  Detail & Related papers  (2026-06-15T11:50:56Z)
• Operational Feature Fingerprints of Graph Datasets via a White-Box Signal-Subspace Probe [0.3058685580689604]
  We propose WG-SRC, a white-box signal-subspace probe for prediction and graph dataset diagnosis.<n>It replaces learned message passing with a fixed, named graph-signal dictionary containing raw features.<n>As a white-box graph-learning instrument, WG-SRC uses predictive performance to validate its diagnostics.
  arXiv  Detail & Related papers  (2026-04-24T16:00:53Z)
• Explainability-Guided Adversarial Attacks on Transformer-Based Malware Detectors Using Control Flow Graphs [0.19116784879310025]
  This paper examines the vulnerability of a RoBERTa-based malware detector that linearizes control flow graphs into sequences of function calls.<n>By evaluating evasion strategies within this graph-to-sequence framework, we provide insight into the practical robustness of transformer-based malware detectors beyond aggregate detection accuracy.
  arXiv  Detail & Related papers  (2026-04-04T19:50:04Z)
• Dynamic analysis enhances issue resolution [53.50448142467294]
  DAIRA (Dynamic Analysis-enhanced Issue Resolution Agent) is an automated repair framework that embeds dynamic analysis into the agent's reasoning cycle.<n> Driven by a Test Tracing-Driven methodology, DAIRA utilizes lightweight monitors to extract critical runtime data.<n>When powered by Gemini 3 Flash Preview, DAIRA establishes a new state-of-the-art (SOTA) performance, achieving a 79.4% resolution rate on the SWE-bench Verified dataset.
  arXiv  Detail & Related papers  (2026-03-23T14:48:54Z)
• HGT-Scheduler: Deep Reinforcement Learning for the Job Shop Scheduling Problem via Heterogeneous Graph Transformers [0.0]
  Job Shop Scheduling Problem (JSSP) is commonly formulated as a disjunctive graph in which nodes represent operations and edges encode technological precedence constraints as well as machine-sharing conflicts.<n>Most existing reinforcement learning approaches model this graph as homogeneous, merging job-precedence and machine-contention edges into a single relation type.<n>We propose the Heterogeneous Graph Transformer (HGT)-Scheduler, a reinforcement learning framework that models the JSSP as a heterogeneous graph.
  arXiv  Detail & Related papers  (2026-03-06T17:59:12Z)
• Explainable Attention-Guided Stacked Graph Neural Networks for Malware Detection [2.6436521007616114]
  We propose a novel stacking ensemble framework for graph-based malware detection and explanation.<n>Our framework improves classification performance while providing insightful interpretations of malware behavior.
  arXiv  Detail & Related papers  (2025-08-13T13:33:02Z)
• Structural Alignment Improves Graph Test-Time Adaptation [17.564393890432193]
  We introduce Test-Time Structural Alignment (TSA), a novel algorithm for Graph Test-Time Adaptation (GTTA)<n>TSA aligns graph structures during inference without accessing the source data.<n>Experiments on synthetic and real-world datasets demonstrate TSA's consistent outperformance of both non-graph TTA methods and state-of-the-art GTTA baselines.
  arXiv  Detail & Related papers  (2025-02-25T16:26:25Z)
• Catch Causal Signals from Edges for Label Imbalance in Graph Classification [2.0316763723596063]
  We leverage edge information to disentangle the causal subgraph from the original graph.<n>Our design leads to improved performance on graph classification tasks with label imbalance issues.<n>We evaluate our approach on real-word datasets PTC, Tox21, and ogbg-molhiv.
  arXiv  Detail & Related papers  (2025-01-03T09:09:58Z)
• Graph Structure Refinement with Energy-based Contrastive Learning [56.957793274727514]
  We introduce an unsupervised method based on a joint of generative training and discriminative training to learn graph structure and representation.<n>We propose an Energy-based Contrastive Learning (ECL) guided Graph Structure Refinement (GSR) framework, denoted as ECL-GSR.<n>ECL-GSR achieves faster training with fewer samples and memories against the leading baseline, highlighting its simplicity and efficiency in downstream tasks.
  arXiv  Detail & Related papers  (2024-12-20T04:05:09Z)
• Learning to Model Graph Structural Information on MLPs via Graph Structure Self-Contrasting [50.181824673039436]
  We propose a Graph Structure Self-Contrasting (GSSC) framework that learns graph structural information without message passing. The proposed framework is based purely on Multi-Layer Perceptrons (MLPs), where the structural information is only implicitly incorporated as prior knowledge. It first applies structural sparsification to remove potentially uninformative or noisy edges in the neighborhood, and then performs structural self-contrasting in the sparsified neighborhood to learn robust node representations.
  arXiv  Detail & Related papers  (2024-09-09T12:56:02Z)
• Graph Signal Sampling for Inductive One-Bit Matrix Completion: a Closed-form Solution [112.3443939502313]
  We propose a unified graph signal sampling framework which enjoys the benefits of graph signal analysis and processing. The key idea is to transform each user's ratings on the items to a function (signal) on the vertices of an item-item graph. For the online setting, we develop a Bayesian extension, i.e., BGS-IMC which considers continuous random Gaussian noise in the graph Fourier domain.
  arXiv  Detail & Related papers  (2023-02-08T08:17:43Z)
• Counterfactual Intervention Feature Transfer for Visible-Infrared Person Re-identification [69.45543438974963]
  We find graph-based methods in the visible-infrared person re-identification task (VI-ReID) suffer from bad generalization because of two issues. The well-trained input features weaken the learning of graph topology, making it not generalized enough during the inference process. We propose a Counterfactual Intervention Feature Transfer (CIFT) method to tackle these problems.
  arXiv  Detail & Related papers  (2022-08-01T16:15:31Z)

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.