A latent mediation framework with sparse autoencoders enables non-additive token-level influence attribution in LLMs by learning orthogonal features and back-propagating attributions.
Neel Nanda, Lawrence Chan, Tom Lieberum, Jess Smith, and Jacob Steinhardt
3 Pith papers cite this work, alongside 8 external citations. Polarity classification is still indexing.
3
Pith papers citing it
8
external citations · external index
fields
cs.LG 3years
2026 3verdicts
UNVERDICTED 3representative citing papers
Grokking delay on Collatz prediction is a decoder access bottleneck after early encoder structure learning, with numeral base as a strong inductive bias on learnability.
Emergent capabilities arise stochastically from abrupt learning of sparse attention patterns on synthetic linear map and cellular automata tasks, with larger models learning them earlier on average.
citing papers explorer
-
Correcting Influence: Unboxing LLM Outputs with Orthogonal Latent Spaces
A latent mediation framework with sparse autoencoders enables non-additive token-level influence attribution in LLMs by learning orthogonal features and back-propagating attributions.
-
The Long Delay to Arithmetic Generalization: When Learned Representations Outrun Behavior
Grokking delay on Collatz prediction is a decoder access bottleneck after early encoder structure learning, with numeral base as a strong inductive bias on learnability.
-
Emergent Capabilities Arise Randomly from Learning Sparse Attention Patterns
Emergent capabilities arise stochastically from abrupt learning of sparse attention patterns on synthetic linear map and cellular automata tasks, with larger models learning them earlier on average.