Run name: real_world_v1
This dataset contains real-world HTML examples for index-based web extraction.
Rows preserve noisy webpage blocks and use teacher-voted interval labels (3-of-5 threshold).
Files:
canonical/real_world_v1.canonical.jsonl: full-page indexed examples (365 rows)
annotated/real_world_v1.with_fireworks.jsonl: teacher annotations (5 passes each)
final/real_world_v1.voted.jsonl: 3-of-5 voted interval labels… See the full description on the dataset page:
https://huggingface.co/datasets/OmAlve/reading-steiner-training-data.
