A long-horizon research agent needs to answer a multi-part question that requires several separate lookups. One design has the model produce an upfront plan listing each lookup it will need and what each one depends on, hand that whole plan to a separate component that executes each lookup, and only then pass the collected results to a final step that composes the answer -- rather than re-invoking the full reasoning model after every single lookup to decide the next step. What is the main advantage of this plan-first, decoupled design over re-reasoning after every individual lookup?
- It is the only design capable of using more than one external tool within a single task
- It guarantees a higher final accuracy than any design that reasons between each lookup, on every possible task
- It eliminates the possibility of any lookup ever failing or returning an unusable result
- It avoids repeatedly re-feeding the full growing transcript of prior reasoning and results back into the expensive reasoning model at every single step, cutting redundant token consumption while remaining robust even if an individual lookup tool fails
Why D? And why not the others?
Correct answer: D. It avoids repeatedly re-feeding the full growing transcript of prior reasoning and results back into the expensive reasoning model at every single step, cutting redundant token consumption while remaining robust even if an individual lookup tool fails
Decoupling planning from execution -- as in the ReWOO approach's Planner, Worker, and Solver split -- lets the reasoning model produce a full blueprint of what lookups are needed just once, hand the actual tool calls to a separate lightweight component, and only bring the reasoning model back in to compose the final answer, instead of re-feeding the whole accumulating transcript of past thoughts and observations into the expensive reasoning model after every single lookup. Xu et al. report this cuts token consumption substantially compared to designs that interleave reasoning and tool calls at every step, while also holding up better when a tool call fails, since the plan does not have to be rebuilt from scratch. The option calling this the only way to use more than one tool is wrong, since interleaved reasoning-and-acting designs also chain multiple tool calls; the difference here is efficiency and structure, not tool count. The option claiming guaranteed higher accuracy on every task is wrong, since decoupling planning from execution can lose the benefit of adjusting the plan mid-task when an early result changes what is actually needed. The option claiming lookups can never fail or return something unusable is wrong -- the design is reported to be more robust to tool failure, not immune to it.
Source: Xu et al., 'ReWOO: Decoupling Reasoning from Observations for Efficient Augmented Language Models' (2023), arXiv:2305.18323