JournalFit-NLP derived evaluation data for evidence-resolved scholarly journal recommendation

This Version 2 dataset contains legally shareable derived evaluation data and reproducibility files from the JournalFit-NLP scholarly journal recommendation project. It includes 152,735 evidence-resolved active submission venue identifiers, frozen title-hash splits, query-level ranks and ranking metrics for six retrieval systems, cross-encoder and local-LLM diagnostic outputs, aggregate and stratified tables, bootstrap confidence intervals, paired randomization tests, protocol and environment records, data dictionaries, checksums, figures, and public validation scripts. The release does not redistribute licensed article abstracts, licensed venue-profile text, vector caches, or QLoRA adapter weights; the excluded artifacts and available hashes are documented in restricted_manifest.json.