KV caches function as notebooks of prefilled conclusions, enabling field-level edits that recover decisions (especially with CoT) and position-portable skill composition with near-identical outputs at O(L) cost.
Mass- editing memory in a transformer
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 4roles
background 1polarities
background 1representative citing papers
SAEs used for layer selection with raw task vectors outperform subspace projection and raise math reasoning accuracy on Gemma-3-4B-IT.
Conflict and hallucination in transformers are basin competition versus basin absence in hidden-state space; geometric margin detects them with zero false refusals while entropy cannot, and confident hallucinations scale as exp(-c/Δ̄).
SCoL trains LLMs via meta-reinforcement learning to generate layer-specific update instructions that improve knowledge acquisition and retention from context streams over standard baselines.
citing papers explorer
-
Models Take Notes at Prefill: KV Cache Can Be Editable and Composable
KV caches function as notebooks of prefilled conclusions, enabling field-level edits that recover decisions (especially with CoT) and position-portable skill composition with near-identical outputs at O(L) cost.
-
Interpretability-Guided Layer Selection over Subspace Projection: SAEs as Stethoscopes, Not Scalpels, for Raw Task Vector Model Editing
SAEs used for layer selection with raw task vectors outperform subspace projection and raise math reasoning accuracy on Gemma-3-4B-IT.
-
Attractor Geometry of Transformer Memory: From Conflict Arbitration to Confident Hallucination
Conflict and hallucination in transformers are basin competition versus basin absence in hidden-state space; geometric margin detects them with zero false refusals while entropy cannot, and confident hallucinations scale as exp(-c/Δ̄).
-
Self-Consolidating Language Models: Continual Knowledge Incorporation from Context
SCoL trains LLMs via meta-reinforcement learning to generate layer-specific update instructions that improve knowledge acquisition and retention from context streams over standard baselines.