Averaging next-token probabilities across paraphrased prompts in a single batched pass gives small but consistent gains on translation, code, and simplification benchmarks.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
M-Ped: Multi-Prompt Ensemble Decoding for Large Language Models
Averaging next-token probabilities across paraphrased prompts in a single batched pass gives small but consistent gains on translation, code, and simplification benchmarks.