humans& open-sources a hardware-native NVFP4 4-bit RL training recipe
gharik · x · 2026-11-14
AI company humans& has open-sourced a hardware-native 4-bit RL recipe that significantly accelerates training. The company trains models from the long-term impacts of their interactions with people, which requires prioritizing long-horizon multi-agent RL.
The author adds technical color: fully 4-bit RL with NVFP4 poses many cool challenges—leveraging lessons across the stack, drawing on NVFP4 pretraining and decode techniques, while minimizing mismatch between trainers and samplers.
More from Infra
- Airtel chairman: Starlink V3 delivers 1 Tbps per launch, matching all of OneWeb — elonmusk · 2026-11-14
- Dev goes all-in on Cloudflare, citing its excellent MCP as the deciding factor — ritakozlov · 2026-11-14
- Brave Leo adds verifiable encrypted LLM inference via TEEs, calls ZDR clauses a pinky promise — gnukeith · 2026-11-14
- 97% of coding-agent tokens are cache reads, but cache writes eat 31% of the cost — Nitjsefnie · 2026-11-14
- workerd lands KV, Queues and Workflows in cluster, first release approaching — irvinebroque · 2026-11-14
- AWS Details GS Caltex MLOps Pipeline: Model Deployment Cut from a Week to Hours — AWS ML Blog · 2026-11-14