# google/uncertainty-baselines

High-quality implementations of standard and SOTA methods on a variety of tasks.

Repository: https://github.com/google/uncertainty-baselines
Canonical: https://ross.abutalabs.com/products/uncertainty-baselines
Language: Python
License: Apache-2.0
License Family: permissive
Topics: bayesian-methods, deep-learning, machine-learning, data-science, tensorflow, neural-networks, statistics, probabilistic-programming
Last push: 2026-08-26T13:45:52+00:00

## Health v2 (maintenance only)
Score: 77/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 35, longevity 100
- inputs: {"age_days": 2240, "days_push": 7, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1592, forks 223 (observed 2026-08-28T04:05:08.712543+00:00)

## What it is
A library of high-quality, minimal-dependency implementations of standard and state-of-the-art uncertainty and robustness methods for deep learning, built on TensorFlow. It serves as a template for researchers to benchmark and prototype new ideas against consistent baselines.

## Use cases
- benchmark uncertainty estimation methods on CIFAR and ImageNet
- reproduce SOTA Bayesian deep learning baselines
- prototype new uncertainty or robustness ideas on top of standard baselines
- compare deterministic vs probabilistic neural network training
- run uncertainty experiments on TPUs via Colab or Google Cloud
- fork a baseline training script for a paper

## When to choose
- you research uncertainty quantification or robustness in deep learning
- you need consistent, comparable baselines for a paper
- you want forkable TensorFlow training scripts with minimal interdependencies
- you need TPU-ready experiment setups

## When to avoid
- you need a stable released API for production
- you work in PyTorch rather than TensorFlow
- you need general-purpose ML tooling unrelated to uncertainty
- you want a maintained pip-installable stable version

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, benchmarking, testing
- domain: machine-learning, deep-learning, data-science
- platform: python, cloud
- tags: uncertainty-quantification, bayesian-deep-learning, tensorflow, robustness, research-baselines, probabilistic-modeling, gpu

## Member repositories
- google/uncertainty-baselines (main) score 77

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:08.712543+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:54:40.431235+00:00, confidence not recorded.
  - readme: https://github.com/google/uncertainty-baselines (fetched 2026-08-28T04:05:08.712543+00:00, sha 3f37f06d68e3)
  - registry_pypi: https://pypi.org/pypi/uncertainty-baselines/json (fetched 2026-08-29T11:25:10.557917+00:00, sha f9ad338395fe)
- Data as of 2026-08-30T08:39:29.467469+00:00.