Stable reinforce handles 8-minute rollouts with FP4 weights in open RL training
simonguozirui · x · 2026-11-06
Commentator ar0cket1 analyzes the Inkling release: stable reinforce (SC) copes well with heavy policy lag even at 8-minute average rollouts, using FP4 weights for training, sampling, and KV cache. He notes SC seems obvious in retrospect and is surprised it hadn't been explored before.
More from Infra
- Zyphra ships ZYFA: FP8 attention on AMD MI355X at 1.16-1.76x AITER throughput — bclavie · 2026-11-06
- Akamai's signed AI book tops $14B, anchored by Anthropic's $11.6B commitment — pdamodaran · 2026-11-06
- Teknium pushes back on METR's 'no bottlenecks' claim, citing funding and compute cost inflation — Teknium · 2026-11-06
- Fudan's D3D decouples diffusion drafting for asynchronous speculative decoding speedups — Fudan-University · 2026-11-06
- ThursdAI pod: midterms, 40 datacenter ballot measures, $3T wipeout — is AI over? — altryne · 2026-11-06
- sshhole Bug Analysis: Alpine/MUSL Systems Hit Hard, ~100-500M OpenWRT Routers Were at Risk — esrtweet · 2026-11-06