REVIEW 7 cited by
Gradient Routing: Masking Gradients to Localize Computation in Neural Networks
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Neural networks are trained primarily based on their inputs and outputs, without regard for their internal mechanisms. These neglected mechanisms determine properties that are critical for safety, like (i) transparency; (ii) the absence of sensitive information or harmful capabilities; and (iii) reliable generalization of goals beyond the training distribution. To address this shortcoming, we introduce gradient routing, a training method that isolates capabilities to specific subregions of a neural network. Gradient routing applies data-dependent, weighted masks to gradients during backpropagation. These masks are supplied by the user in order to configure which parameters are updated by which data points. We show that gradient routing can be used to (1) learn representations which are partitioned in an interpretable way; (2) enable robust unlearning via ablation of a pre-specified network subregion; and (3) achieve scalable oversight of a reinforcement learner by localizing modules responsible for different behaviors. Throughout, we find that gradient routing localizes capabilities even when applied to a limited, ad-hoc subset of the data. We conclude that the approach holds promise for challenging, real-world applications where quality data are scarce.
Forward citations
Cited by 7 Pith papers
-
Train the Model, Not the Reader: Decodability Supervision for Verifiable Activation Explanations
Reconstruction scores do not certify individual claims in activation explanations; co-adapted private codes can carry the score, and target-side training (RECAP) makes designated content verifiably decodable.
-
Modular Pretraining Enables Access Control
GRAM selectively trains auxiliary modules so that ablating one at inference removes a targeted capability while preserving the rest, closely tracking data-filtered models at 5x lower cost across 5 capability profiles.
-
Policy-Masked Private Experts: Auditable and Reversible Capability Access Control in Sparse MoE Models
A hard mask applied before top-k MoE routing lets trusted authorization choose between a frozen public and a newly trained private expert pool, giving auditable zero private execution under deny.
-
Studying Cross-cluster Modularity in Neural Networks
A clusterability regularizer creates strongly separated clusters in neural networks, shrinking effective circuit size by up to 90% on CIFAR-10, but it does not create task-specialized modules.
-
Physics of Skill Learning
The paper introduces Geometry, Resource, and Domino models that reproduce the sequential Domino effect in skill learning and link it to scaling laws, optimizers, and modularity.
-
Preserving Bilinear Weight Spectra with a Signed and Shrunk Quadratic Activation Function
SQS is a new GLU activation that reportedly preserves bilinear weight spectra for interpretability while achieving competitive performance on MNIST, FMNIST, and Tiny Stories.
-
Perspectives for Direct Interpretability in Multi-Agent Deep Reinforcement Learning
A perspective paper that advocates direct, post hoc interpretability for multi-agent deep reinforcement learning and offers a taxonomy of where those methods might apply.
Discussion (0). Continue with ORCID to comment.