DynamixSFT adaptively reweights instruction-tuning datasets during training using a prior-anchored bandit and one-step loss-reduction reward, giving small but consistent benchmark gains.
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
CONDITIONAL 1representative citing papers
citing papers explorer
-
DynamixSFT: Dynamic Mixture Optimization of Instruction Tuning Collections
DynamixSFT adaptively reweights instruction-tuning datasets during training using a prior-anchored bandit and one-step loss-reduction reward, giving small but consistent benchmark gains.