NLFT weights each token by how much its probability shifts under different natural language prompts and claims to beat supervised fine-tuning with 50 examples, but the paper's loss equation and reported gains are internally inconsistent.
Reinforcement learning from hu- man feedback research program,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2024 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Natural Language Fine-Tuning
NLFT weights each token by how much its probability shifts under different natural language prompts and claims to beat supervised fine-tuning with 50 examples, but the paper's loss equation and reported gains are internally inconsistent.