Tencent Hunyuan extends critical-batch-size theory to LLM RL: 29% faster GRPO, 2.29× PPO throughput
TencentHunyuan · x · 2026-09-24
Tencent Hunyuan's new research extends classical critical-batch-size theory to online LLM RL, where the model generates its own training data and rollout scaling differs from training scaling. Key findings: across GRPO and PPO, learning-rate retuning preserves learning per response over a bounded batch-size range; on fixed hardware, larger batches boost PPO generation-stage throughput by up to 2.29×; and the best GRPO configuration hits the same validation target in 29% less time.
More from Infra
- Gas-for-AI-datacenters forum co-located with CoRL shows energy demand spillover — carlosdponx · 2026-11-12
- a16z podcast: Mike Volpi on Cisco's 70+ acquisition playbook and the AI buildout — a16z · 2026-11-12
- AMD to pass Intel in server revenue by mid-2027 as share gains 6 points a year — BenBajarin · 2026-11-12
- Study: Reflective Data Center Roofs Actually Cool Daytime Ground by 1.8-2.2°C — SumitGup · 2026-11-12
- Tsinghua's BaGuaLu-2 tames MoE expert-load spikes at runtime, accepted to SC26 — jiqizhixin · 2026-11-12
- antirez backs low-bit DeepSeek 4.1 Flash over Qwen 27B dense on Apple M5 Macs — antirez · 2026-11-12