Unverified50% confidenceBenchmarkExact time
On LLaDA-2.1-mini, ORL requires only 4.2M self-generated training tokens and 36 H100 GPU hours of post-training to obtain up to a 3x throughput improvement with comparable quality.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/17/2026
First Seen
Valid until: 10/31/2026
Sources
Related Entities
Related Claims
UnverifiedTraining a multi-task policy using Diffusion Policy on 8 A100 GPUs typically takes 24–72 hours.74% similarUnverifiedTraining uses batch size 256, 400,000 steps, EMA 0.9999, linear learning rate decay over the final quarter, and torch.compile achieves a 2.4× speedup over eager mode.72% similarUnverifiedMeta's Llama 2 (70-billion-parameter version) used 1.8 million GPU hours for training.71% similarUnverifiedThe 20.6-second generation time does not include training cost; training used 16 V100 GPUs.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1049948API
curl https://kongchang.com/api/v1/knowledge/claims/1049948MCP
get_claim(id=1049948)