Injecting the logit gap between a small RL-trained model and its base into a larger base model at decoding time improves reasoning accuracy on math and code benchmarks, sometimes matching or exceeding RL training of the large model.
SFT memorizes, RL generalizes: A comparative study of foundation model post-training
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
RAST: Reasoning Activation in LLMs via Small-model Transfer
Injecting the logit gap between a small RL-trained model and its base into a larger base model at decoding time improves reasoning accuracy on math and code benchmarks, sometimes matching or exceeding RL training of the large model.