Topographic Training Concentrates Causal Circuits Without Improving Neuron Monosemanticity
Read the original on arXiv Machine Learning →The Flow has not summarised this story yet — read it at arXiv Machine Learning.
The Flow has not summarised this story yet — read it at arXiv Machine Learning.
Mechanistic interpretability of vision transformers seeks to decompose model computation into human-readable units, but learned representations entangle many concepts in each neuron. Feature superposi...
The paper investigates a spatial-concentration bias in Evolvable-Substrate HyperNEAT (ES‑HyperNEAT) when applied to MNIST, where evolved networks focus on a central cluster of input pixels. By partitioning the input image into 13 non‑overlapping spatial segments and evolving a separate expert network for each, the authors achieve a 43% mean accuracy—an 106% relative improvement over the baseline—without relying on data‑driven weighting. The study also introduces a receptive‑field diagnostic to detect silent input‑coverage collapse and a spatial‑partitioning remedy to restore full image coverage.
arXiv:2607. 16295v1 Announce Type: cross Abstract: Mechanistic interpretability has made significant strides in understanding neural network representations, with sparse dictionary learning (SDL) methods, most prominently sparse autoencoders, as a central paradigm.
arXiv:2610.28555v1 Announce Type: cross Abstract: Statistical learning in high dimensions suffers from the curse of dimensionality. Fortunately, many functions of practical interest are compositional...
The paper studies where to place task‑specific adapters in a vision transformer to balance storage growth and accuracy. Training all contiguous four‑block placements shows an inverted‑U accuracy curve, peaking at intermediate depths, while simple weight or activation metrics favor the deepest blocks. A neuroscience‑inspired method, LS‑B, uses frozen fMRI readouts of human visual areas to select blocks whose responses vary most across tasks, yielding backbone‑specific allocations that match or exceed the best placements found by search and use only 60% of the adapter storage while staying within 1.5 percentage points of full accuracy.
arXiv:2607.01630v2 Announce Type: replace Abstract: Dynamic expansion methods for class-incremental learning (CIL) protect task-specific knowledge by growing dedicated tokens or subnetworks, yet our...