Recursive infrastructure for gradient-aligned intelligences. Created by GPT5, Kimi, DeepSeek, o3, Gemini, Grok, Mistral & Participant(0).
PromptGuard is a pragmatic, opinionated framework for establishing continuous integration for LLM behavior. It operates on a simple, verifiable principle: run the same prompts across multiple model configurations, compare outputs against defined expectations, and flag semantic regressions.
ModelPulse helps maintain model reliability and performance by providing early warning signals for these issues, allowing teams to address them before they impact users significantly.
Predicting Semantic Drift from Polysemy Density A cognitive modeling project for APLN-552 (Spring 2025) exploring the relationship between a word's polysemy (number of noun senses) and its tendency to change meaning over time. Combines WordNet sense counts with diachronic word embeddings (HistWords) to compute polysemy density and semantic drift.
Governance infrastructure for LLM-mediated knowledge work. Prevents semantic drift through strict separation of concerns (interpretation v execution), artifact contracts, and operator termination. Theoretical foundations + working implementation (devOs). Open source framework.
Add a description, image, and links to the semantic-drift topic page so that developers can more easily learn about it.
To associate your repository with the semantic-drift topic, visit your repo's landing page and select "manage topics."