Fine-tuning a small encoder-decoder T5 model with hierarchical tax-code tokens and constrained beam search improves HSN/SAC code prediction over flat classifiers and other SLM architectures in the authors' internal benchmark.
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
-
Domain-Adaptive Small Language Models for Structured Tax Code Prediction
Fine-tuning a small encoder-decoder T5 model with hierarchical tax-code tokens and constrained beam search improves HSN/SAC code prediction over flat classifiers and other SLM architectures in the authors' internal benchmark.