Standard circuit discovery methods produce dataset-specific circuits rather than task-general ones, and a new clustering-based method discovers multiple more faithful circuits per dataset.
Title resolution pending
7 Pith papers cite this work. Polarity classification is still indexing.
7
Pith papers citing it
citation-role summary
background 2
citation-polarity summary
verdicts
UNVERDICTED 7roles
background 2polarities
background 2representative citing papers
LLMs implement both attention-based suppression and constructive representations for negation, with construction dominant, despite poor accuracy from late-layer attention shortcuts.
Circuit-based metrics from Vision Transformer internals provide better label-free proxies for generalization under distribution shift than existing methods like model confidence.
MoE expert neurons show lower polysemanticity than dense FFN neurons, widening with sparser routing, and experts specialize in fine-grained tasks like specific linguistic operations, supporting expert-level interpretability.
A new framework using Task Subspace Logit Attribution localizes attention heads specialized for task recognition and task learning in in-context learning, showing they align and rotate hidden states within a task subspace.
Sparse autoencoders plus greedy filtering and factorization-machine interaction modeling identify minimal sets of features in Gemma-2-2B-IT and LLaMA-3.1-8B-IT whose ablation produces jailbreaks by flipping refusal to compliance.
Per-head attention contributions to the residual stream serve as strong linear features for classifying relational knowledge in LLMs, with probe accuracy correlating to relation specificity and signal distribution.
citing papers explorer
-
Data-driven Circuit Discovery for Interpretability of Language Models
Standard circuit discovery methods produce dataset-specific circuits rather than task-general ones, and a new clustering-based method discovers multiple more faithful circuits per dataset.
-
How Language Models Process Negation
LLMs implement both attention-based suppression and constructive representations for negation, with construction dominant, despite poor accuracy from late-layer attention shortcuts.
-
Inside-Out: Measuring Generalization in Vision Transformers Through Inner Workings
Circuit-based metrics from Vision Transformer internals provide better label-free proxies for generalization under distribution shift than existing methods like model confidence.
-
The Expert Strikes Back: Interpreting Mixture-of-Experts Language Models at Expert Level
MoE expert neurons show lower polysemanticity than dense FFN neurons, widening with sparser routing, and experts specialize in fine-grained tasks like specific linguistic operations, supporting expert-level interpretability.
-
Localizing Task Recognition and Task Learning in In-Context Learning via Attention Head Analysis
A new framework using Task Subspace Logit Attribution localizes attention heads specialized for task recognition and task learning in in-context learning, showing they align and rotate hidden states within a task subspace.
-
Beyond I'm Sorry, I Can't: Dissecting Large Language Model Refusal
Sparse autoencoders plus greedy filtering and factorization-machine interaction modeling identify minimal sets of features in Gemma-2-2B-IT and LLaMA-3.1-8B-IT whose ablation produces jailbreaks by flipping refusal to compliance.
-
Tracing Relational Knowledge Recall in Large Language Models
Per-head attention contributions to the residual stream serve as strong linear features for classifying relational knowledge in LLMs, with probe accuracy correlating to relation specificity and signal distribution.