Introduces a benchmark using logical rules from knowledge graphs to generate multi-hop questions that evaluate whether knowledge edits in LLMs propagate to entailed facts, finding up to 24% performance gaps for methods like ROME and FT.
Preprint, arXiv:2502.11196
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
verdicts
UNVERDICTED 3roles
background 1polarities
background 1representative citing papers
Circuit-based metrics from Vision Transformer internals provide better label-free proxies for generalization under distribution shift than existing methods like model confidence.
Sparse crosscoders on LLM checkpoint triplets track emergence, maintenance, and discontinuation of linguistic features during pretraining via a new RelIE metric.
citing papers explorer
-
Benchmarking Knowledge Editing using Logical Rules
Introduces a benchmark using logical rules from knowledge graphs to generate multi-hop questions that evaluate whether knowledge edits in LLMs propagate to entailed facts, finding up to 24% performance gaps for methods like ROME and FT.
-
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.
-
Crosscoding Through Time: Tracking Emergence & Consolidation Of Linguistic Representations Throughout LLM Pretraining
Sparse crosscoders on LLM checkpoint triplets track emergence, maintenance, and discontinuation of linguistic features during pretraining via a new RelIE metric.