Themis AI BlogResearch and engineering notes on uncertainty quantification, risk-aware machine learning, and trustworthy AI from the Themis AI team.https://www.themisai.io/Spotting the Unfamiliar: Out-of-Distribution Detection with Uncertaintyhttps://www.themisai.io/research/out-of-distribution-detection/https://www.themisai.io/research/out-of-distribution-detection/Models fail most dangerously on inputs they've never seen. Epistemic uncertainty flags those out-of-distribution cases in real time — so the system can escalate, abstain, or adapt.Wed, 20 May 2026 00:00:00 GMTManaging Risk in Production AI Systemshttps://www.themisai.io/research/managing-risk-in-production-ai/https://www.themisai.io/research/managing-risk-in-production-ai/Why production AI needs risk measurement, not just predictions — and how to turn a model's uncertainty into enforceable risk policies.Wed, 15 Apr 2026 00:00:00 GMTNIST's Risk Management Framework and Themis AIhttps://www.themisai.io/research/nists-risk-management-framework-and-evidential/https://www.themisai.io/research/nists-risk-management-framework-and-evidential/Examining the Risk Management Framework by NIST.Tue, 16 Sep 2025 00:00:00 GMTCatching LLM Hallucinations with Uncertaintyhttps://www.themisai.io/research/catching-llm-hallucinations/https://www.themisai.io/research/catching-llm-hallucinations/LLMs fail without warning. A separate uncertainty estimate lets a model abstain, escalate, or defer instead of confidently making things up.Tue, 18 Feb 2025 00:00:00 GMTDrug Discovery Cost Reduction with Uncertainty-Guided Predictionshttps://www.themisai.io/research/drug-discovery-and-clinical-trials/https://www.themisai.io/research/drug-discovery-and-clinical-trials/Drug Discovery Cost Reduction with Uncertainty-Guided PredictionsTue, 03 Dec 2024 00:00:00 GMTUncertainty-Aware Human Intervention for Autonomous Vehicleshttps://www.themisai.io/research/autonomous-driving/https://www.themisai.io/research/autonomous-driving/Uncertainty-Aware Human Intervention for Autonomous Vehicles.Thu, 05 Sep 2024 00:00:00 GMTRisk-Aware Hallucination Detection for Arbitrary Generative Modelshttps://www.themisai.io/research/generative-models/https://www.themisai.io/research/generative-models/Risk-Aware Hallucination Detection for Arbitrary Generative Models.Thu, 20 Jun 2024 00:00:00 GMTUncertainty-Aware Language Modeling for Selective Question Answeringhttps://www.themisai.io/research/uncertainty-aware-language-modeling/https://www.themisai.io/research/uncertainty-aware-language-modeling/Presenting uncertainty-aware LLMs capable of estimating uncertainty with every prediction.Sun, 26 Nov 2023 00:00:00 GMTProbability vs Confidencehttps://www.themisai.io/research/probability-vs-confidence/https://www.themisai.io/research/probability-vs-confidence/At Themis AI, we are building a groundbreaking technology to create and ensure trustworthy and robust AI.Wed, 13 Sep 2023 00:00:00 GMTRobust and Trustworthy Deep Learning (part 3): Themis AIhttps://www.themisai.io/research/robust-and-trustworthy-deep-learning-part-3-themis-ai/https://www.themisai.io/research/robust-and-trustworthy-deep-learning-part-3-themis-ai/Themis AI's cutting-edge technological advancements in robust and trustworthy deep learningFri, 25 Aug 2023 00:00:00 GMTRobust and Trustworthy Deep Learning (Part 2): Uncertaintyhttps://www.themisai.io/research/robust-and-trustworthy-deep-learning-part-2-uncertainty/https://www.themisai.io/research/robust-and-trustworthy-deep-learning-part-2-uncertainty/Themis AI's cutting-edge technological advancements in robust and trustworthy deep learningFri, 18 Aug 2023 00:00:00 GMTRobust and Trustworthy Deep Learning (Part 1): Biashttps://www.themisai.io/research/robust-and-trustworthy-deep-learning-part-1/https://www.themisai.io/research/robust-and-trustworthy-deep-learning-part-1/Themis AI's cutting-edge technological advancements in robust and trustworthy deep learningFri, 11 Aug 2023 00:00:00 GMTPreliminary Steps Towards Risk-Aware Image Generationhttps://www.themisai.io/research/stable-diffusion/https://www.themisai.io/research/stable-diffusion/Using Capsa to automatically evaluate the quality of images generated by Stable DiffusionFri, 28 Jul 2023 00:00:00 GMT