Two high schoolers evolve s-means, beating Meta's e-means with ~55% lower error
IgorCarron · x · 2026-11-16
Two high school students used their EvoForest evolutionary search (9 island-evolution rounds) to improve Meta's e-means clustering into s-means, which matches e-means' 8-epoch quality in just 1-3 epochs with 55% lower error on GPUs. They shipped fused Triton CUDA kernels, PyTorch fallbacks and an open-source scikit-learn-compatible package, with benchmarks on T4, L4 and A100.
More from Research
- Continual-ARC: lifelong ARC-AGI stream finds agent harness persistence rarely beats simple caching — unifyai · 2026-11-16
- DySENSE: MIT's GenAI framework synthesizes satellite imagery with dynamic city energy maps — MIT · 2026-11-16
- WebArena-Pro: a 300-task multimodal benchmark across 20 self-hosted apps for web agents — McGill-NLP · 2026-11-16
- Training-free page-sparse decode in llama.cpp: 1.72× faster 128k decoding on an RTX 5090 — topppggg · 2026-11-16
- Training Qwen3-8B to drive a simulated hand spawns a dedicated internal 'pressure' direction — repligate · 2026-11-16
- Google Research open-sources PaperOrchestra, a multi-agent framework for writing AI papers — google-research · 2026-11-16