A reward-oriented influence scoring method that uses pairwise preference loss to select 5% of instruction-tuning data, outperforming similarity-based selection baselines on SHP, SE, and HH-RLHF.
Free dolly: Introducing the world’s first truly open instruction-tuned llm
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
ROSE: A Reward-Oriented Data Selection Framework for LLM Task-Specific Instruction Tuning
A reward-oriented influence scoring method that uses pairwise preference loss to select 5% of instruction-tuning data, outperforming similarity-based selection baselines on SHP, SE, and HH-RLHF.