cff-version: 1.2.0 message: >- If you use ASSERT, please cite it as below. Please also consider citing the prior work it builds on (see the references section). title: "ASSERT: Adaptive Spec-driven Scoring for Evaluation and Regression Testing" abstract: >- ASSERT is an open-source framework for turning natural-language behavior specifications into executable evaluations for AI models, applications, and agents. type: software authors: - name: "Microsoft Responsible AI" repository-code: "https://github.com/responsibleai/ASSERT" url: "https://github.com/responsibleai/ASSERT" license: MIT references: - type: article title: "AI-Assisted Systematization for Evaluating GenAI Systems" year: 2026 month: 5 url: "https://www.microsoft.com/en-us/research/publication/ai-assisted-systematization-for-evaluating-genai-systems/" notes: >- ASSERT's concept-systematization stage follows this work. authors: - family-names: Agarwal given-names: Dhruv - family-names: Sheng given-names: Emily - family-names: Atalla given-names: Chad - family-names: Garcia-Gathright given-names: Jean - family-names: Mozannar given-names: Hussein - family-names: Washington given-names: Hannah - family-names: Chouldechova given-names: Alexandra - family-names: Barocas given-names: Solon - family-names: Wallach given-names: Hanna - type: software title: "Bloom: an open source tool for automated behavioral evaluations" year: 2025 url: "https://github.com/safety-research/bloom" notes: >- ASSERT's staged evaluation pipeline is modeled on Bloom's design. Bloom's MIT license notice is retained in THIRD_PARTY_NOTICES.md as good-faith attribution. authors: - family-names: Gupta given-names: Isha - family-names: Fronsdal given-names: Kai - family-names: Sheshadri given-names: Abhay - family-names: Michala given-names: Jonathan - family-names: Tay given-names: Jacqueline - family-names: Wang given-names: Rowan - family-names: Bowman given-names: "Samuel R." - family-names: Price given-names: Sara