CommonLit Readability Prize

Rate the complexity of literary passages for grades 3-12 to improve reading level assessment.

NLP RMSE BiGRU + Ridge

Feature Importance

text_embedding 0.35 sentence_count 0.15 word_count 0.12 avg_word_len 0.10 syllable_count 0.10 flesch_score 0.08 text_length 0.05

Key Insights

External data + teacher/student framework wins. Sentence transformers + DeBERTa ensemble. Small dataset benefits greatly from knowledge distillation.