Skip to content

pqid-bench v1.1.0

Release date: 2026-07-30

Scope

Version 1.1.0 is a software and documentation release for the frozen PQID-Bench v1.0.0 study. It does not change any benchmark prompt, split, response, target, evaluator result, structural predicate, model roster, denominator, or reported numerical finding. The benchmark archive remains identified by DOI 10.5281/zenodo.21649753.

Added

  • pqid-bench dashboard, which creates a standalone interactive Plotly report from the frozen evidence bundle;
  • an importable visualization API: load_dashboard_data, build_dashboard, and write_site_assets;
  • five coordinated evidence views covering the measurement ladder, model profiles, component recovery, stochastic repeatability, and provider-route aggregates;
  • an accessible tabular representation of every plotted model row;
  • a GitHub Pages documentation workflow that generates presentation assets after validating the source evidence tree;
  • a navigable MkDocs Material manual, workflow diagram, and static README fallback; and
  • a visualization optional dependency group based on Plotly 6.

Boundaries

The explorer is a presentation layer. It neither contacts a model provider nor executes generated code. Fresh provider collection remains the responsibility of pqid-bench run-model; executable evaluation remains isolated in the Docker-backed pqid-bench replay path.

Interactive HTML and generated SVG assets are Pages build products. They are not inserted into the DOI-bound frozen evidence archive or the Hugging Face dataset payload. PyPI renders the static README and links to the hosted explorer because package indexes do not execute embedded JavaScript.

Version Dimensions

Dimension Identifier
Python package 1.1.0
benchmark release 1.0.0
evaluator pqid-bench-evaluator-1.1.0-safe-builtins
structural predicate pqid-bench-reference-signature-1.0.0-count-map
schema 1.0.0
frozen artifact manifest 1.0.0
container artifact 1.0.0