On GPT-2 hash tasks, LoRA layers improved a synthetic reasoning task more than a planning task, and the authors interpret this as evidence that reasoning is inherently low rank.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2024 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Planning vs Reasoning: Ablations to Test Capabilities of LoRA layers
On GPT-2 hash tasks, LoRA layers improved a synthetic reasoning task more than a planning task, and the authors interpret this as evidence that reasoning is inherently low rank.