Repository containing code for the paper: Fixing Overconfidence in Dynamic Neural Networks
-
Updated
Dec 7, 2023 - Python
Repository containing code for the paper: Fixing Overconfidence in Dynamic Neural Networks
A library for calibrating and visualizing the Softmax function for probabilistic interpretation.
Safety before certainty: an open-source public-data stress-test showing a structured evidence-sufficiency prompt cuts unsafe overconfident clinical answers across four frontier LLMs (Opus 4.8, GPT-5.5, Gemini 3.5 Flash, Grok 4.3), with matched circularity controls and blinded clinician judge-validation.
Measuring and fixing DNN overconfidence on MNIST with ECE and temperature scaling.
AI agent metacognition skills distilled from David Dunning's Self-Insight (2005): calibration, Dunning–Kruger awareness, the outside view, and feedback discipline for Claude Code, Codex, and compatible AI agents.
Almost everyone is overconfident, and almost no one keeps score. Weather forecasters and professional superforecasters are well calibrated for exactly one reason: they write predictions down with probabilities and get graded. This app gives you the same loop with zero ceremony.
Add a description, image, and links to the overconfidence topic page so that developers can more easily learn about it.
To associate your repository with the overconfidence topic, visit your repo's landing page and select "manage topics."