Memorized facts in fine-tuned LLMs often sit off the mid-layer reasoning path; relocating those representations recovers most multi-hop generalization failures.
Cake: Circuit-aware editing enables generalizable knowledge learners
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.AI 2years
2026 2representative citing papers
Power-law data distributions outperform uniform ones for compositional reasoning by creating asymmetry that lets frequent skill compositions scaffold rare ones with less data.
citing papers explorer
-
Towards Mechanistically Understanding Why Memorized Knowledge Fails to Generalize in Large Language Model Finetuning
Memorized facts in fine-tuned LLMs often sit off the mid-layer reasoning path; relocating those representations recovers most multi-hop generalization failures.
-
The Power of Power Law: Asymmetry Enables Compositional Reasoning
Power-law data distributions outperform uniform ones for compositional reasoning by creating asymmetry that lets frequent skill compositions scaffold rare ones with less data.