The Best Instruction-Tuning Data are Those That Fit

NeurIPSSpotlight2025

Authors
Dylan Zhang, Qirun Dai, Hao Peng
Venue
NeurIPS 2025
Track
Spotlight

TL;DR

High-quality supervised finetuning (SFT) data are essential for unlocking pretrained LLMs’ capabilities…

Opening excerpt from the authors’ abstract. source

Read the paper

Topics

llm

Related NeurIPS papers

← All NeurIPS 2025 Spotlight papers · Browse the whole archive