Tencent Hunyuan Research Shows Larger Batches Speed Up LLM Reinforcement Learning

Tencent's Hunyuan team extends critical-batch-size theory to online LLM RL, showing larger batches cut time to target performance when hardware throughput gains outweigh sample-efficiency losses, boosting PPO training throughput up to 2.29x and GRPO by 29%.

2026-09-24 ~ 2026-09-25 · 3 related posts

• Tencent Hunyuan extends critical-batch-size theory to LLM RL: 29% faster GRPO, 2.29× PPO throughput — TencentHunyuan · 2026-09-24
• Hunyuan Research: Batch-Size Scaling with LR Retuning Boosts PPO Throughput 2.29x — TencentHunyuan · 2026-09-24
• Tencent RL study: larger batches speed up LLM RL if throughput gains beat sample-efficiency loss — CurieuxExplorer · 2026-09-25