Entity · technique

Cross-sample Consistency Regularization

techniqueactiveprovisionalcross-sample-consistency-regularization-3d0ab1af·1 events·first seen Jun 30, 2026

Aliases: Cross-sample Consistency Regularization

Co-occurring entities

Sparse Autoencoders C²R: Cross-sample Consistency Regularization Mitigates Feature Splitting and Absorption in Sparse Autoencoders

More like this (12)

C²R: Cross-sample Consistency Regularization Mitigates Feature Splitting and Absorption in Sparse Autoencoders R-Drop consistency regularization KL-Cov regularization Entropy Regularization consistency training L0 regularization Entropy-Regularized Reinforcement Learning Semantic Edit Regularization Dynamic-Probabilistic Consistency Gap Divergence Regularized Policy Optimization KL-regularized RL Consistency Training Can Entrench Misalignment

Recent events (1)

5arXiv · cs.LG·Jun 30, 2026·source ↗

C²R regularization method addresses feature splitting and absorption in Sparse Autoencoders

A new arXiv preprint introduces C²R (Cross-sample Consistency Regularization), a training technique for Sparse Autoencoders (SAEs) that mitigates two known failure modes: feature splitting, where coherent concepts fragment across multiple latents, and feature absorption, where general features develop arbitrary exceptions. C²R penalizes co-activation of directionally similar latents across a batch, encouraging each semantic concept to map consistently to a single latent. The authors report that C²R reduces both pathologies while preserving reconstruction fidelity, with source code released publicly.

Evaluation and Benchmarking AI Safety Research Sparse Autoencoders Cross-sample Consistency Regularization C²R: Cross-sample Consistency Regularization Mitigates Feature Splitting and Absorption in Sparse Autoencoders