Long-Horizon Agent
An AI agent that must make tens or more consecutive decision steps. Its reasoning chain is long, it is sensitive to small per-step errors, and errors accumulate and amplify along the chain.
Timeline (last 90 days)
The core bottlenecks of long-horizon agent training are the scarcity of training data and the lack of training methods, with reward signals being extremely sparse.
The rollout length of long-horizon agents has been extended to several hours and on the order of nearly a million tokens.
All Facts (2)
The core bottlenecks of long-horizon agent training are the scarcity of training data and the lack of training methods, with reward signals being extremely sparse.
50%UnverifiedThe rollout length of long-horizon agents has been extended to several hours and on the order of nearly a million tokens.
50%