CLG selects many-shot demonstrations by matching fine-tuning gradients of a small language model to the full training set, improving accuracy over random selection by 2-4%.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Selecting Demonstrations for Many-Shot In-Context Learning via Gradient Matching
CLG selects many-shot demonstrations by matching fine-tuning gradients of a small language model to the full training set, improving accuracy over random selection by 2-4%.