Which approach breaks a large goal into smaller tasks before acting?
Why B is the answer
Plan-and-execute
A planner creates the steps, and executors carry them out.
In detail
- Plan-and-execute creates the steps, then carries them out.
- Replan when things change.
Use case
An illustrative scenario, not a real incident.
A planner breaks a goal into tasks, and executors carry them out.
See it in different ways
stateDiagram-v2 [*] --> Reason Reason --> Act Act --> Observe Observe --> Reason: not done Observe --> [*]: done
mindmap
root((Memory))
Short term
Current context
Summaries
Long term
Episodic
Semantic
Planning
Plan and execute
Revise on new factsflowchart LR S0["Goal"] --> S1["Planner"] --> S2["Tasks"] --> S3["Executors"]
Key point: Plan first, then execute.
Why the other options are wrong
- A. Mixed precision
- That saves training memory.
- C. Top-k sampling
- Top-k is sampling.
- D. Gradient checkpointing
- That saves training memory.