Dissecting Jet-Tagger Through Mechanistic Interpretability

• URL: http://arxiv.org/abs/2605.09881v1
• Date: Mon, 11 May 2026 02:11:47 GMT
• Title: Dissecting Jet-Tagger Through Mechanistic Interpretability
• Abstract summary: We perform a mechanistic interpretability analysis of the Particle Transformer architecture, trained on the Top Quark Tagging reference dataset.<n>We identify the computational circuit responsible for jet classification and characterizing the physical content of its internal representations.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Mechanistic interpretability seeks to reverse engineer a trained neural network by identifying the minimal subset of internal components. We perform a mechanistic interpretability analysis of the Particle Transformer architecture, trained on the Top Quark Tagging reference dataset, with the goal of identifying the computational circuit responsible for jet classification and characterizing the physical content of its internal representations. Combining zero ablation, path patching with two complementary on-manifold corruption strategies and linear probing of the residual stream, we identify a sparse six-head circuit that recovers the great majority of the full model performance while admitting a clean source-relay-readout interpretation. In this circuit, a single early layer head serves as the primary causal source, a cluster of middle-layer heads acts as relays selectively attending to hard pairwise substructure and a single late-layer head reads out the aggregated signal. Linear probes show that the residual stream is preferentially aligned with the energy correlator basis over the $N$-subjettiness basis. Within the energy correlator basis, the model preferentially encodes 2-prong substructure observables over the 3-prong observables. A per-layer trained probe further reveals that the apparent single step commitment of the model to a classification decision in the first class attention block is in fact a basis rotation, with the discriminating signal already saturating in the particle attention stack. These results demonstrate that mechanistic interpretability methods developed for natural language models can be used for jet physics classifiers and indicate that gradient descent may rediscover physically meaningful aspects of jet tagging without supervision.

Related papers

• Finding and using interpretable latents in a neutrino foundation model with sparse autoencoders [3.3461260868741527]
  We present a first application of sparse-autoencoder-based mechanistic interpretability to particle physics.<n>We study a neutrino foundation model pretrained on IceCube data and fine-tuned for direction reconstruction.<n>Motivated by this underused information, we train an uncertainty head on the same event-level representation to predict the model's angular reconstruction error.
  arXiv  Detail & Related papers  (2026-08-26T17:53:00Z)
• Seeing Through Circuits: Faithful Mechanistic Interpretability for Vision Transformers [54.2202951543048]
  We investigate whether useful mechanistic circuits can be identified through computational graphs in vision transformers.<n>We propose an effective method for Automatic Visual Circuit Discovery (Vi-CD) that recovers class-specific circuits for classification.<n>We find that insightful and actionable edge-based circuits can be recovered from vision transformers, adding transparency to the internal computations of these models.
  arXiv  Detail & Related papers  (2026-04-15T23:25:46Z)
• Weight Patching: Toward Source-Level Mechanistic Localization in LLMs [12.558998651017374]
  We propose Weight Patching, a parameter-space intervention method for source-oriented analysis in paired same-architecture models.<n>We introduce a framework centered on a vector-anchor behavioral interface that provides a shared internal criterion for whether a task-relevant control state has been formed.
  arXiv  Detail & Related papers  (2026-04-15T10:21:38Z)
• Towards foundation-style models for energy-frontier heterogeneous neutrino detectors via self-supervised pre-training [0.4151704012621244]
  We present a sparse ViT framework for learning reusable representations from heterogeneous detector data.<n>Self-supervised pre-training combines masked autoencoder reconstruction with relational voxel-level objectives.<n>We find that pre-training consistently improves neutrino flavour and charm-quark identification.
  arXiv  Detail & Related papers  (2026-04-08T12:52:00Z)
• TopoCurate:Modeling Interaction Topology for Tool-Use Agent Training [53.93696896939915]
  Training tool-use agents typically rely on Supervised Fine-Tuning (SFT) on successful trajectories and Reinforcement Learning (RL) on pass-rate-selected tasks.<n>We propose TopoCurate, an interaction-aware framework that projects multi-trial rollouts from the same task into a unified semantic quotient topology.<n>TopoCurate achieves consistent gains of 4.2% (SFT) and 6.9% (RL) over state-of-the-art baselines.
  arXiv  Detail & Related papers  (2026-03-02T10:38:54Z)
• How Attention Sinks Emerge in Large Language Models: An Interpretability Perspective [67.08306259095778]
  We identify a simple mechanism that enables the model to recognize token at position zero and induce an attention sink within two transformer blocks.<n>By analyzing training traces from a 30B A3B MoE model trained from scratch, we find that this mechanism emerges early in training and becomes increasingly concentrated in the first two layers.
  arXiv  Detail & Related papers  (2026-02-04T11:10:34Z)
• Disentangling Recall and Reasoning in Transformer Models through Layer-wise Attention and Activation Analysis [3.1526281887627587]
  Distinguishing recall from reasoning is crucial for predicting model generalization.<n>We use controlled datasets of synthetic linguistic puzzles to probe transformer models at the layer, head, and neuron level.<n>Our results provide the first causal evidence that recall and reasoning rely on separable but interacting circuits in transformer models.
  arXiv  Detail & Related papers  (2025-10-03T04:13:06Z)
• Learning local discrete features in explainable-by-design convolutional neural networks [0.0]
  We introduce an explainable-by-design convolutional neural network (CNN) based on the lateral inhibition mechanism. The model consists of the predictor, that is a high-accuracy CNN with residual or dense skip connections. By collecting observations and directly calculating probabilities, we can explain causal relationships between motifs of adjacent levels.
  arXiv  Detail & Related papers  (2024-10-31T18:39:41Z)
• SINDER: Repairing the Singular Defects of DINOv2 [61.98878352956125]
  Vision Transformer models trained on large-scale datasets often exhibit artifacts in the patch token they extract. We propose a novel fine-tuning smooth regularization that rectifies structural deficiencies using only a small dataset.
  arXiv  Detail & Related papers  (2024-07-23T20:34:23Z)
• Unsupervised Interpretable Basis Extraction for Concept-Based Visual Explanations [44.033369364364084]
  This work attempts to explain CNN image classifier predictions and intermediate layer representations in terms of human-understandable concepts.<n>We take a bottom-up approach, identifying the directions from the structure of the feature space, collectively, without relying on supervision from concept labels.<n>We make extensions to existing basis interpretability metrics and show that intermediate layer representations become more interpretable when transformed with the extracted bases.
  arXiv  Detail & Related papers  (2023-03-19T00:37:19Z)
• Recurrent Spectral Network (RSN): shaping the basin of attraction of a discrete map to reach automated classification [4.724825031148412]
  A novel strategy to automated classification is introduced which exploits a fully trained dynamical system to steer items toward distinct attractors. Non-linear terms act for a transient and allow to disentangle the data supplied as initial condition to the discrete dynamical system. Our novel approach to classification, that we here term Recurrent Spectral Network (RSN), is successfully challenged against a simple test-bed model, created for illustrative purposes, as well as a standard dataset for image processing training.
  arXiv  Detail & Related papers  (2022-02-09T14:59:06Z)
• Data-driven emergence of convolutional structure in neural networks [83.4920717252233]
  We show how fully-connected neural networks solving a discrimination task can learn a convolutional structure directly from their inputs. By carefully designing data models, we show that the emergence of this pattern is triggered by the non-Gaussian, higher-order local structure of the inputs.
  arXiv  Detail & Related papers  (2022-02-01T17:11:13Z)
• BiDet: An Efficient Binarized Object Detector [96.19708396510894]
  We propose a binarized neural network learning method called BiDet for efficient object detection. Our BiDet fully utilizes the representational capacity of the binary neural networks for object detection by redundancy removal. Our method outperforms the state-of-the-art binary neural networks by a sizable margin.
  arXiv  Detail & Related papers  (2020-03-09T08:16:16Z)

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.