# Apple — Plan or Retrieve First? A Scale Study of Tool Discovery Architectures

- Company: Apple (apple.com)
- Announced: 2026-10-15
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://machinelearning.apple.com/research/plan-retrieve-tool-discovery
- Record: https://forck.live/items/23074-plan-or-retrieve-first-a-scale-study-of-tool-discovery-architectures
- Subject: Machine Learning Research

Apple researchers studied two tool discovery architectures for LLM agents: Retrieve-then-Plan (RtP) and Plan-then-Retrieve (PtR). Evaluating on ToolBench across single-tool and multi-tool queries, they found a complexity-dependent crossover where RtP outperforms PtR on single-tool queries while PtR outperforms RtP on multi-tool queries. The crossover holds across retrieval budgets, registry sizes, and model families, and the paper characterizes a stub-hallucination failure mode that acts as a double-edged sword.

## Evidence

Verbatim from https://machinelearning.apple.com/research/plan-retrieve-tool-discovery:

> We study an alternative Plan-then-Retrieve (PtR) paradigm in which the LLM first generates Abstract Tool Stubs (hypothesized function signatures describing needed capabilities) which then serve as retrieval queries against the registry.

---

Record: https://forck.live/items/23074-plan-or-retrieve-first-a-scale-study-of-tool-discovery-architectures
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md