# Can AI Models Direct Each Other? Organizational Structure as a Probe into Training Limitations

arXiv: https://arxiv.org/abs/2603.26458 · HTML: https://mlsanity.com/paper/2603.26458/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 16 of its 17 references that are in the map.
- First posted on arXiv 27 Mar 2026.

## Builds on (most cited first, 16 of 16)

- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation](https://mlsanity.com/paper/2308.08155/) (2023)
- [MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework](https://mlsanity.com/paper/2308.00352/) (2023)
- [SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering](https://mlsanity.com/paper/2405.15793/) (2024)
- [OpenHands: An Open Platform for AI Software Developers as Generalist Agents](https://mlsanity.com/paper/2407.16741/) (2024)
- [ChatDev: Communicative Agents for Software Development](https://mlsanity.com/paper/2307.07924/) (2023)
- [AgentVerse: Facilitating Multi-Agent Collaboration and Exploring Emergent Behaviors](https://mlsanity.com/paper/2308.10848/) (2023)
- [Agentless: Demystifying LLM-based Software Engineering Agents](https://mlsanity.com/paper/2407.01489/) (2024)
- [AutoCodeRover: Autonomous Program Improvement](https://mlsanity.com/paper/2404.05427/) (2024)
- [Towards a Science of Scaling Agent Systems](https://mlsanity.com/paper/2512.08296/) (2025)
- [SWE-Search: Enhancing Software Agents with Monte Carlo Tree Search and Iterative Refinement](https://mlsanity.com/paper/2410.20285/) (2024)
- [Self-planning Code Generation with Large Language Models](https://mlsanity.com/paper/2303.06689/) (2023)
- [Live-SWE-agent: Can Software Engineering Agents Self-Evolve on the Fly?](https://mlsanity.com/paper/2511.13646/) (2025)
- [HyperAgent: Generalist Software Engineering Agents to Solve Coding Tasks at Scale](https://mlsanity.com/paper/2409.16299/) (2024)
- [BudgetMLAgent: A Cost-Effective LLM Multi-Agent system for Automating Machine Learning Tasks](https://mlsanity.com/paper/2411.07464/) (2024)
- [Orchestrating Human-AI Teams: The Manager Agent as a Unifying Research Challenge](https://mlsanity.com/paper/2510.02557/) (2025)

## Built on by (most cited first, 2 of 2)

- [LoopArena: Benchmarking Models as Runtime Controllers for Loop Engineering](https://mlsanity.com/paper/2608.28281/) (2026)
- [Scaling Inference-Time Compute with Teams of Coding Agents](https://mlsanity.com/paper/2610.17471/) (2026)

## Abstract (arXiv)

Can an expensive AI model effectively direct a cheap one to solve software engineering tasks? We study this question by introducing ManagerWorker, a two-agent pipeline where an expensive "manager" model (text-only, no code execution) analyzes issues, dispatches exploration tasks, and reviews implementations, while a cheap "worker" model (with full repo access) executes code changes. We evaluate on 200 instances from SWE-bench Lite across five configurations that vary the manager-worker relationship, pipeline complexity, and model pairing. Our findings reveal both the promise and the limits of multi-agent direction: (1) a strong manager directing a weak worker (62%) matches a strong single agent (60%) at a fraction of the strong-model token usage, showing that expensive reasoning can substitute for expensive execution; (2) a weak manager directing a weak worker (42%) performs worse than the weak agent alone (44%), demonstrating that the directing relationship requires a genuine capability gap--structure without substance is pure overhead; (3) the manager's value lies in directing, not merely reviewing--a minimal review-only loop adds just 2pp over the baseline, while structured exploration and planning add 11pp, showing that active direction is what makes the capability gap productive; and (4) these behaviors trace to a single root cause: current models are trained as monolithic agents, and splitting them into director/worker roles fights their training distribution. The pipeline succeeds by designing around this mismatch--keeping each model close to its trained mode (text generation for the manager, tool use for the worker) and externalizing organizational structure to code. This diagnosis points to concrete training gaps: delegation, scoped execution, and mode switching are skills absent from current training data.