Model Stitching

Follow this topic Ask about this topic

Latest papers

9 papers2 papers in the last four weeks, with none the four weeks before. 0.0% of all new papers.

What the Primary Lund Plane Omits: A Model-Stitching Diagnosis of a Jet-Tagger Representation Gap
Oct 27, 2026Saurabh Rai, Sanmay GangulyModel Stitching Jet Tagging
Department of Physics, Indian Institute of Technology Kanpur, Uttar Pradesh 208016, India
Train to Stitch: Topology-Preserving Regularization for Interoperable Representations
Oct 14, 2026Akshay Malhotra, Nicolas Zilberstein, Shahab Hamidi-RadModel Stitching Latent Space Alignment
InterDigital Communications Inc., Los Altos, CA, USA · Rice University, Houston, TX, USA
Unsupervised Latent Space Alignment with Hyperspherical Geodesic Matching
Aug 28, 2026Cameron Ryan, Vivek Sivaraman Narayanaswamy, Kowshik Thopalli +1Model Stitching Latent Space Alignment
Northeastern University, Boston, MA 02115. Work performed during a summer internship at Lawrence Livermore National Laboratory · Lawrence Livermore National Laboratory, Livermore, CA 94551-0808.
A Stitch in Time Saves Nine: Preserving Policy Compatibility Under Perception Updates in End-to-End Autonomous Driving
Jun 19, 2026Yueyuan Li, Yifei Xiao, Mingyang Jiang +3End-to-End Autonomous Driving Autonomous Driving Perception
School of Automation and Intelligent Sensing, Shanghai Jiao Tong University, Key Laboratory of System Control and Information Processing, Ministry of Education of China, Shanghai, 200240, CN · School of Aeronautics and Astronautics, Shanghai Jiao Tong University, Shanghai, 200240, CN · Global Institute of Future Technology, Shanghai Jiao Tong University, Shanghai, 200240, CN
Forgetting is Not Erasure: Recovering Latent Knowledge via Transport Keys
Jun 1, 2026Archie ChaudhuryContinual Learning Representation Alignment
Axionic Labs
Stitched Value Model for Diffusion Alignment
May 19, 2026Hyojun Go, Hyungjin Chung, Prune Truong +8Diffusion Model Alignment Diffusion Model Guidance
1ETH Zurich · 2Google · University of Copenhagen
Revisiting Model Stitching In the Foundation Model Era
Mar 12, 2026Zheda Mai, Ke Zhang, Fu-En Wang +6Model Stitching Vision Foundation Models
1The Ohio State University · 3Amazon · 2Boston University
Multi-DNN Inference of Sparse Models on Edge SoCs
Mar 10, 2026Jiawei Luo, Di Wu, Simon Dobson +1Model Stitching Efficient Neural Network Inference
School of Computer Science, University of St Andrews, Scotland UK
Connecting Neural Models Latent Geometries with Relative Geodesic Representations
Jun 2, 2025Hanlin Yu, Berfin Inal, Georgios Arvanitidis +3Neural Representation Geometry Latent Space Geometry
University of Helsinki · University of Amsterdam · DTU +1

Oct 27, 2026J/K move · Enter open · S save

What the Primary Lund Plane Omits: A Model-Stitching Diagnosis of a Jet-Tagger Representation Gap

Saurabh Rai, Sanmay Ganguly

Department of Physics, Indian Institute of Technology Kanpur, Uttar Pradesh 208016, India

Model Stitching Jet Tagging

Open paper AskPDF arXiv

Lund-plane and Particle Transformer (ParT) jet taggers can differ in performance because one representation is weaker, or because a weaker downstream computation reads an adequate one. Standalone performance cannot separate these. Model stitching can: it links the frozen early layers of one network to the frozen late layers of another through a single trained linear map. As a positive control we use an incomplete Lund representation, the primary declustering branch, whose deficit is known independently. Stitched into ParT's downstream, it reaches a background rejection at least as much as its own downstream recovers, yet far below ParT's background rejection: the deficit is representational, not computational. A hypergraph reader over splitting triplets does no better, ruling out higher-order correlations as the cause. The gap is invisible in AUC, which stays within 1% of its target value, and appears only in background rejection at low signal efficiency. It localizes to 0.6% of background jets, at and above the top mass, with more constituents but fewer primary splittings than typical background. Rebuilding the representation on the full Cambridge/Aachen declustering tree repairs the deficit as predicted: standalone rejection rises, closing 72% of the gap. Masking nodes at inference shows the repaired network still relies almost entirely on the primary branch; the secondary splittings make the representation learnable rather than supplying information it reads. Along the primary branch, the widest-angle splitting, near the top and

W

opening angles, is worth nearly twice a random one. For quark/gluon tagging the primary branch is far less deficient, retaining 77.6% of ParT's rejection (46.9% for top), and the full tree closes only 33% of the gap (72% for top).