Pinned Loading

1. autoeit autoeit Public

   Whisper transcription + automated 0–4 scoring for Spanish elicited-imitation tests. Rule-based vs LLM scorer agreement: quadratic-weighted κ = 0.71.

   Python

2. osuda osuda Public

   Length-calibrated psychology assistant: QLoRA fine-tune of LLaMA 3.1 8B, served over Telegram. ROUGE-1 0.23->0.44, BERTScore F1 0.86->0.91.

   Jupyter Notebook