twangodev/radiotalk-us-transcripts-grok-4.20-50k
radiotalk-us-transcripts-grok-4.20-50k 49,984 synthetic US air-traffic-control transcripts, generated with xAI's grok-4.20-0309-non-reasoning against the v2 radiotalk scenario pipeline. Built for fine-tuning ATC ASR models (NVIDIA Parakeet, Whisper, etc.) and for seeding TTS audio generation. Third release in the radiotalk transcripts series, and the first from a non-Qwen generator: v1: twangodev/radiotalk-us-transcripts-qwen3-100k v2:… See the full description on the dataset page: https://huggingface.co/datasets/twangodev/radiotalk-us-transcripts-grok-4.20-50k.
docs: attach quality evals (judge scores, STT WER)
eval: per-transcript LLM-judge realism scores
docs: dataset card
data: 49,984 grok-4.20 transcripts (v3, seed 42, deduped)
initial commit