Prime Intellect's AnomalyXL trains a 4B recursive model to SoTA on long time-series anomaly detection
HankYeomans · x · 2026-10-15
- Prime Intellect RL resident @nzuma0 built AnomalyXL, an infinitely scalable synthetic time-series generator with programmatic ground truth, written in verifiers — filling a gap since almost no open RL environments cover long-context time series.
- He trained a 4B TimeRLM on it with prime-rl, generating fresh data every step; the model recursively calls sub-agents via nano-rlm to localize anomalies across 131k datapoints.
- The accompanying paper reports SoTA performance on long-context anomaly localization, with zero-shot transfer to real-world ECG, sleep, and software observability data.
More from Research
- 2021 slides reveal Garçon, Anthropic's internal interpretability infra for multi-GPU models — voooooogel · 2026-10-15
- Meta taps Muse models to build AI research scaffold on 5M health records — _jasonwei · 2026-10-15
- Microduck robot runs onboard SLAM with just 64 pixels of depth, 4.8 cm median error — chris_j_paxton · 2026-10-15
- UIUC paper explains sparse LLM post-training updates through the lens of Fisher information — tokenbender · 2026-10-15
- Shannon's 1971 recollection: von Neumann suggested 'entropy' because 'no one really knows what entropy really is' — eigenron · 2026-10-15
- Two-person team trains HobbyLM MoE in ~304 GPU-hours, 100-400x less compute than peers — msharmas · 2026-10-15