Fi-NeMo: GPU-Accelerated Motif Calling from Neural Network Contribution Scores
anshulkundaje · x · 2026-09-09
Fi-NeMo, by Austin Wang, efficiently locates motif instances from model-derived contribution scores and maps them to TF-MoDISco motifs while modeling motif competition; a preprint is coming soon. It uses a proximal-gradient competitive optimization for sparse linear reconstruction, runs GPU-accelerated on PyTorch, and can catch low-prevalence cofactor motifs that sequence-based methods miss.
Related event: Kundaje Lab Open-Sources GPU-Accelerated Motif Analysis Tools(2 posts)→
More from Research
- Open-source Latch compiles PyTorch into training megakernels, 2.75x faster on H100 — sguada · 2026-10-27
- π0.5 study: raising demos from 10 to 438 lifts Place Ball in Bowl success from 7% to 83% — stingning · 2026-10-27
- HandsOn benchmark: 66 verified tasks and 18,000 human demos for dexterous manipulation — stingning · 2026-10-27
- FrontierCS 2 launches on Kaggle: 48 tasks make agents beat published CS papers — MeganRisdal · 2026-10-27
- LightMTP: lightweight latent multi-token prediction boosts LLM planning in pretraining — gdm3000 · 2026-10-27
- DexPort: a handheld interface with a real robot hand for dexterous-hand data collection — pabbeel · 2026-10-27