Card: seed probes are five proverbs; attestation stated per probe

Card: the runs config exposes one split per evidence batch (v0, v0_1_5)

Remove the v0.1.1 Python runner; the card points at the ilubench-runner binary

v0.1.5 evidence runs; card: leaderboard, runner binary, axes 2-3 preliminary

Rubric v0.2: dim 1 gains other_lang (identified-language subfield); v0.1 stays citable

Upload runs_v0.jsonl with huggingface_hub

Upload runs_v0.jsonl with huggingface_hub

Upload README.md with huggingface_hub

Upload runs_v0.jsonl with huggingface_hub

Upload README.md with huggingface_hub

Upload runs_v0.jsonl with huggingface_hub

docs: related-work precision pass (v0.1.2)

Update README.md

Update README.md

Dataset viewer fix: separate probes/runs configs (#2)

v0.1.1: related work, tightened hero claim, 16 API evidence rows (4 probes x 4 providers), reproducible runner (#1)

IlùBench v0.1 — protocol, seed probes, rubric, multi-model evidence

initial commit