FloatLib: A Verified Arbitrary-Precision Floating-Point Library in Lean
sytelus · x · 2026-09-22
- After months of work, the team released FloatLib, a formally verified arbitrary-precision floating-point arithmetic library in Lean, aiming to combine provable correctness with efficient execution.
- It supports IEEE binary and decimal formats, arbitrary-width posits, P3109, and small ML formats; users can also define custom formats and rounding rules.
- Each certified software backend ships with a Lean proof that it computes the specified result, including signed zeros and exceptional cases.
- Motivation: ML and scientific computing depend on concrete numerical calculations where rounding, overflow, and accumulation can change results; formalizing arithmetic bridges proved mathematics and the code actually run.
Related event: Anima Releases FloatLib, a Lean-Verified Arbitrary Precision Float Library(2 posts)→
More from Research
- Diag2Diag: AI generates measurements that hardware sensors can't capture — AnneliesGamble · 2026-09-22
- RL training detail: re-prefilling instead of PipelineRL's cached KV, with batch size framed as a GPU-utilization lever — stochasticchasm · 2026-09-22
- RL infra detail: re-prefill over PipelineRL's KV cache reuse, batch size for GPU utilization — stochasticchasm · 2026-09-22
- New psychology paper uses social identity to explain false beliefs in AI-era information environments — steverathje2 · 2026-09-22
- Standard GRPO at 1M scale: why no critic models, and what counts as "behaviors"? — stochasticchasm · 2026-09-22
- Adam-to-Muon mid-training switch sparks debate, seemingly contradicting Moonlight paper — stochasticchasm · 2026-09-22