A fixed logit-bias vector estimated from black-box rollouts gives small accuracy gains and modest length compression, far below LoRA fine-tuning.
Note on the general case of the bayes-laplace formula for inductive or a posteriori probabilities.Transactions of the Faculty of Actuaries, 8(1):182–192, 1920
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Frustratingly Simple Black-Box Adaptation of Language Models via Logit Bias
A fixed logit-bias vector estimated from black-box rollouts gives small accuracy gains and modest length compression, far below LoRA fine-tuning.