New Soft-Label MNIST Testbed Captures Human Confidence on Digit Recognition

Annotators labeled each MNIST and Mukhoti digit with yes, no, or unsure, producing image-level soft labels and uncertainty scores for comparing human and model confidence.

Read the full story at Towards Data Science ↗

Timeline · 1 report

2026-11-12 13:30 · Towards Data Science
A Soft-Label MNIST Testbed for Human vs. Model Confidence

More stories

Jensen Huang resigns as Nvidia CEO after espionage and affair reports — CSET Georgetown
Google DeepMind launches AlphaGenome 2 to predict genetic variant effects by cell type — Google DeepMind Blog
Google DeepMind's AlphaProof Code formally verifies Signal's cryptographic code in four weeks — Google DeepMind Blog
United Airlines blocks AI agents from booking flights on customers' behalf — Computerworld AI
Google launches Tabletop, an experimental AI platform for group planning and collaboration — Google AI Blog
Google launches Gemini Decision 1 model and Decisions API for fast closed-set choices — Google Cloud AI Blog
Google expands Gemini 4 Carbon access to Pro plan users in AI Mode — Google DeepMind Blog
AI chip startup Fractile raises $720M at $7B valuation — Techmeme

Get the daily brief of stories like this at 6:30 every morning →