CinderD/wildtrace
WildTrace strict481 WildTrace is a source-internal long-context multi-hop reasoning benchmark built from natural evidence trails. Unlike reverse-synthetic QA, its tasks are mined in situ from long source documents before questions are written. The strict481 release contains 481 locked tasks over 214 public long-form sources, with full-document, evidence-withheld evaluation. The model under test receives only the source document and the public question; evidence spans, clue… See the full description on the dataset page: https://huggingface.co/datasets/CinderD/wildtrace.
Fix evaluation harness and normalize rubrics
Add Gemini-native judge compatibility to eval harness
Clean license and corpus header metadata in WildTrace strict481
Clean WildTrace strict481 public release
Clarify license scope: CC-BY covers annotations/scripts only; source texts retain PD/§105/Project-Gutenberg terms + US-jurisdiction note
Add per-source provenance manifest (PD/§105/PG basis per corpus file)
Release WildTrace strict481 public package
initial commit