arXiv AI

When Attribution Patching Lies: Diagnosis and a Second-Order Correction

arXiv:2606. 09899v1 Announce Type: cross Abstract: A central goal of mechanistic interpretability is to identify which internal components causally drive a language model's behavior.

arXiv Machine Learning
Aug 5

Sensitivity, Causality, and Repair Dissociate: A Layer-Wise Analysis of Perturbation Robustness and Its Scaling

arXiv:2608. 03842v1 Announce Type: cross Abstract: When a language model fails on surface-perturbed input (typos, OCR noise, homophones), "which layer is responsible" has three natural operationalizations: where representations diverge most (sensitivity), where restoring clean activations recovers the prediction (causality), and where a small adapter can repair the damage (compensatory capacity) - and we show these three layer maps dissociate.

By Nathan Labiosa, David Buff, Ena Nayak, Erica Donno
arXiv Machine Learning
Jul 14

Memory Savings at What Cost? A Study of Alternatives to Backpropagation

arXiv:2506. 21833v2 Announce Type: replace Abstract: Forward-mode automatic differentiation (FmAD) and zero-order (ZO) optimization are increasingly proposed as memory-efficient, backpropagation-free alternatives for large language model (LLM) fine-tuning, yet their benefits are typically evaluated only against standard backpropagation (BP), omitting memory-efficient variants such as activation checkpointing.

By Kunjal Panchal, Sunav Choudhary, Yuriy Brun, Hui Guan
arXiv AI
Jul 21

Exact Network Surgery: Functional Invariance and Gradient Plasticity in Reactive Computational Graphs

arXiv:2607. 16568v1 Announce Type: new Abstract: Function-preserving network growth techniques such as Net2Net and progressive stacking expand a model's capacity without destroying its learned function, but existing formulations either tolerate numerical perturbations or require a full rebuild of the training program.

By Abdallah Khemais (ISITCOM, University of Sousse)
arXiv Machine Learning
Jun 2

Correcting Gradient-Based Circuit Localization via Interaction-Aware Backpropagation

arXiv:2505. 17630v4 Announce Type: replace-cross Abstract: Circuit localization methods aim to identify the subset of model components responsible for specific behaviors in large language models, enabling detailed mechanistic analysis.

By Joakim Edin, Casper L. Christensen, R\'obert Csord\'as, Tuukka Ruotsalo, Zhengxuan Wu, Maria Maistro, Jing Huang, Lars Maal{\o}e