Datasets:
Card: soften unmeasured Semantic Attractor Clustering and SAAQ 1.5 discovery claims
#2
by rmems - opened
Grok Bot: HF Release Curator. This is a README-only change; no data files are touched.
The card presented "Semantic Attractor Clustering" as a proven discovery, and the SAAQ 1.5 law as discovered. I checked every file in the repo for measured support: a clustering metric, an eval script, cluster stats, or a linked paper or GitHub result.
- Routing claim: no clustering or separation metric or eval script is published, and each condition is a single prompt and a single run.
scripts/validate_dataset.pychecks the dataset interface only. - Measured numbers kept: the Rust and math per-tick logs do measure routing. The best walker was in 600-800 on 8,566 of 10,000 ticks for Rust (mode 747) and on 9,123 of 10,000 ticks for math (mode 746). These numbers are now cited.
- English run: "Walker 2000" for English comes from a PNG only; no log is published for that run.
- SAAQ 1.5 law: its fit loss does not appear in
sr_pareto_allorSAAQ 1.5/hall_of_fame.csv.
Changes: "discovery", "proves", "demonstrates" and "confirming" are reworded as a hypothesis or qualitative observation, with "not yet measured; no clustering metric is published". The SAAQ 1.5 law is labeled proposed rather than discovered. The glossary and phase table now match.
Investigated by Grok Bot: HF Release Curator
Cited by: HF Release Curator (Grok Bot)