TestBED: open-source toolkit for benchmarking Bayesian experimental design
tom_rainforth · x · 2026-11-05
Tom Rainforth's team releases TestBED, an open-source toolkit for Bayesian experimental design. It provides a flexible, extendable foundation for coding adaptive design methods and quickly benchmarking them. Built with Gavin Kerrigan, Marcel Hedman, and Sahel Iqbal; code and docs are on GitHub.
More from Research
- Shared expert design compared across GLM-5.3, Kimi K3, DeepSeek V4.1 Flash — Nouamanetazi · 2026-11-06
- Kimi K3 swaps sign updates for quantile balancing across its 896 experts — Nouamanetazi · 2026-11-06
- DeepSeek-V4's first layers route by token-id hash, no learned router — Nouamanetazi · 2026-11-06
- DeepSeek-V4 switches router scoring to sqrt(softplus); MoE scaling factors diverge — Nouamanetazi · 2026-11-06
- MoE Training Handbook Ch.2: A Runbook From Hardware Tests to Your First MoE — Nouamanetazi · 2026-11-06
- 500K GPU-hours MoE Handbook: 18 of Top 20 Open-Weights Models Are MoEs — Nouamanetazi · 2026-11-06