Part-of-speech template patterns in a distilled student model's outputs can identify its teacher above chance in a closed set of five candidate models, outperforming n-gram and similarity baselines on most tasks.
https:// github.com/tatsu-lab/stanford_alpaca
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
-
Who Taught You That? Tracing Teachers in Model Distillation
Part-of-speech template patterns in a distilled student model's outputs can identify its teacher above chance in a closed set of five candidate models, outperforming n-gram and similarity baselines on most tasks.