Unverified50% confidenceFactExact time
The blue-side main agent was trained to update 45,600; the red-side exploiter was trained for only about 2 hours (half of a 4-hour cycle), with the sole task of defeating a frozen copy of the main agent at step 43,800.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/2/2026
First Seen
Valid until: 1/31/2027
Sources
Related Entities
Related Claims
UnverifiedIn the first round of the PPO reinforcement learning task, Opus chose to train a quadruped ant for nearly an hour, 75 million steps, scoring 7254; Kimi chose to train a bipedal cheetah for less than six minutes, barely passing with 3000 points.73% similarUnverifiedA single Mario training run of 25 million steps requires more than two days just for the simulation phase.71% 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.71% similarUnverifiedThe experiment trains a small MLP on a modular addition task modulo 97, with the training set making up only 35%, using full-batch training for a total of 20k steps.70% similarUnverified仅调整 num_workers、pin_memory、prefetch_factor 三个 DataLoader 参数，相同 2000 步训练耗时从 633 秒降至 358 秒，运行时间缩短 43%69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1125547API
curl https://kongchang.com/api/v1/knowledge/claims/1125547MCP
get_claim(id=1125547)