LENSLLM fits an NTK-augmented scaling law to small fine-tuning subsets to rank candidate LLMs, reporting up to 85.8% ranking correlation, up to 91.1% relative accuracy, and up to 88.5% compute reduction compared to full tuning.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
LENSLLM: Unveiling Fine-Tuning Dynamics for LLM Selection
LENSLLM fits an NTK-augmented scaling law to small fine-tuning subsets to rank candidate LLMs, reporting up to 85.8% ranking correlation, up to 91.1% relative accuracy, and up to 88.5% compute reduction compared to full tuning.