CHIMERA-Hash Ultra: Extending Chaos-IDF with Negation Detection and Factual Variation Handling via Corpus-Free Statistical Signals

CHIMERA-Hash Ultra extends CHIMERA-Hash v1 (DOI: 10.5281/zenodo.18823652) with five additional statistical signals that resolve two documented failure modes: negation detection and factual variation handling. All signals are derived purely from character statistics and set arithmetic — no word lists, no trained models, no corpora. Key results on 115-pair benchmark across 16 categories:- #1 overall (Pearson correlation 0.6940 vs TF-IDF 0.5680)- Negation MAE: 0.1404 (v1 was 0.6249 — 77.5% improvement)- Factual Variation MAE: 0.0561 (v1 was 0.2907 — 80.7% improvement)- 9/16 category wins Novel mechanisms: Short-Alpha-Unique Ratio (SAUR) for negation detection without word lists; LCS + Numeric Jaccard Cap for factual variation detection.