[KongchangAI]
Unverified50% confidenceFactExact time

In traditional on-policy RL, policy updates must wait for an entire trajectory to be sampled, which brings severe throughput bottlenecks in long-horizon tasks.

1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen

Sources

Related Entities

Related Claims

Cite This Claim

Stable URI
https://kongchang.com/claim/1041835
API
curl https://kongchang.com/api/v1/knowledge/claims/1041835
MCP
get_claim(id=1041835)