nondeterminism

Hillel Wayne's newsletter covers his move to NYC, an upcoming Antithesis webinar on property-based testing for web applications, and a technical exploration of the challenges of differential testing on nondeterministic programs, including hyperproperties and distribution-level comparisons.

该论文系统研究了文本分类器在不同服务上下文（批量大小、数值精度、硬件等）下的非不变性，发现标签稳定并不意味着分数稳定，且在 bf16 下仅改变批量形状就能导致高达 56.7 个百分点的预测概率质量偏移，并为每种机制推导了标签稳定性的充分条件和缓解方法。

A developer recounts a three-week production bug where a regex with a literal backspace character silently disabled a language-detection guardrail, making the LLM appear flaky. The post highlights the need to instrument deterministic guardrails to distinguish them from model nondeterminism.

This paper proposes a distribution-aware training approach for modeling next-event predictions in concurrent Go programs, treating scheduler nondeterminism as a signal. Fine-tuning a 7B model on fewer than a thousand traces achieves 36.2% accuracy on production bugs, outperforming Gemini 3.5 Flash zero-shot.

This tweet discusses the idea of training models with 'implementation noise' to improve robustness against float numerics problems caused by nondeterminism and nonassociativity.