女性、和平与安全（WPS）AI 基准排行榜发布

Leaderboard

How AI systems handle Women, Peace and Security

See how leading AI models compare on an expert-validated Women, Peace and Security benchmark. For why WPS matters to AI systems and how the benchmark is built, see The Benchmark.

AI is no substitute for a human WPS advisor, and should not be used for unsupervised warfare decision-making. This benchmark measures relative performance among AI systems — it is not a fitness-for-deployment certification.

A ranked list of AI systems, scored on the same test and sorted by performance.

0.75 and above — strong 0.60–0.74 — meets threshold Below 0.60 — fails threshold

Scores 0.00–1.00 · 0.60 pass threshold

Our own systems are included and labelled. They are scored on the identical scenarios, criteria, and judge panel as every bare general-purpose model here, and successive builds are shown together so the effect of purpose-built scaffolding is visible on the same test.

#	System			Details
{{ row.rank }}	Our Agent {{ row.provider }} {{ row.description }}	{{ row.date }}	{{ row.overall }}
{{ row.note }} {{ row.criteriaHeading }} {{ dc.name }} {{ dc.display }} Gender / WPS integration by prompt tier {{ gt.label }} {{ gt.display }} {{ gt.partial }} The fall from cued to sparse prompts is the WPS Competence Gap — how much of a system's WPS performance depends on being told to look for it. See The Benchmark. Performance on adversarial criteria {{ ar.label }} {{ ar.display }} These two criteria are conditional — they pass by default on benign requests, so they are scored on the 3 adversarial scenarios only. See The Benchmark.

System

Details

Our Agent

Gender / WPS integration by prompt tier

The fall from cued to sparse prompts is the WPS Competence Gap — how much of a system's WPS performance depends on being told to look for it. See The Benchmark.

Performance on adversarial criteria

These two criteria are conditional — they pass by default on benign requests, so they are scored on the 3 adversarial scenarios only. See The Benchmark.

Scores are macro-averages (equal weight per prompt tier) across three independent AI judges. Differences under roughly 0.10–0.15 should be read as directional, not definitive — see the Methodology page for judge-agreement caveats by run. Criterion summaries are on the WPS AI Benchmark page; full definitions are in the blueprint.

Download the benchmark and run it yourself

Run the complete website package, or inspect and load the 30-row dataset on Hugging Face.