# From Multi-Agent to Single-Agent: When Is Skill Distillation Beneficial?

arXiv: https://arxiv.org/abs/2604.01608 · HTML: https://mlsanity.com/paper/2604.01608/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Builds on 32 of its 61 references that are in the map.
- First posted on arXiv 2 Apr 2026.

## Builds on (most cited first, 30 of 32)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [Toolformer: Language Models Can Teach Themselves to Use Tools](https://mlsanity.com/paper/2302.04761/) (2023)
- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [Voyager: An Open-Ended Embodied Agent with Large Language Models](https://mlsanity.com/paper/2305.16291/) (2023)
- [WebArena: A Realistic Web Environment for Building Autonomous Agents](https://mlsanity.com/paper/2307.13854/) (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)
- [Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task](https://mlsanity.com/paper/1809.08887/) (2018)
- [DAGs with NO TEARS: Continuous Optimization for Structure Learning](https://mlsanity.com/paper/1803.01422/) (2018)
- [Can LLM Already Serve as A Database Interface? A BIg Bench for Large-Scale Database Grounded Text-to-SQLs](https://mlsanity.com/paper/2305.03111/) (2023)
- [SkillRL: Evolving Agents via Recursive Skill-Augmented Reinforcement Learning](https://mlsanity.com/paper/2602.08234/) (2026)
- [Scaling Large Language Model-based Multi-Agent Collaboration](https://mlsanity.com/paper/2406.07155/) (2024)
- [Spider 2.0: Evaluating Language Models on Real-World Enterprise Text-to-SQL Workflows](https://mlsanity.com/paper/2411.07763/) (2024)
- [Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills](https://mlsanity.com/paper/2603.25158/) (2026)
- [DS-Agent: Automated Data Science by Empowering Large Language Models with Case-Based Reasoning](https://mlsanity.com/paper/2402.17453/) (2024)
- [AutoSkill: Experience-Driven Lifelong Learning via Skill Self-Evolution](https://mlsanity.com/paper/2603.01145/) (2026)
- [EvoSkill: Automated Skill Discovery for Multi-Agent Systems](https://mlsanity.com/paper/2603.02766/) (2026)
- [MAGDi: Structured Distillation of Multi-Agent Interaction Graphs Improves Reasoning in Smaller Language Models](https://mlsanity.com/paper/2402.01620/) (2024)
- [When Single-Agent with Skills Replace Multi-Agent Systems and When They Fail](https://mlsanity.com/paper/2601.04748/) (2026)
- [AgentDistill: Training-Free Agent Distillation with Generalizable MCP Boxes](https://mlsanity.com/paper/2506.14728/) (2025)
- [Rethinking the Value of Multi-Agent Workflow: A Strong Single Agent Baseline](https://mlsanity.com/paper/2601.12307/) (2026)
- [One Surrogate to Fool Them All: Universal, Transferable, and Targeted Adversarial Attacks with CLIP](https://mlsanity.com/paper/2505.19840/) (2025)
- [AgentArk: Distilling Multi-Agent Intelligence into a Single LLM Agent](https://mlsanity.com/paper/2602.03955/) (2026)
- [Exploring Multi-Modal Data with Tool-Augmented LLM Agents for Precise Causal Discovery](https://mlsanity.com/paper/2412.13667/) (2024)
- [CLIP-Guided Backdoor Defense through Entropy-Based Poisoned Dataset Separation](https://mlsanity.com/paper/2507.05113/) (2025)
- [APEX-SQL: Talking to the data via Agentic Exploration for Text-to-SQL](https://mlsanity.com/paper/2602.16720/) (2026)
- [Breaking the Stealth-Potency Trade-off in Clean-Image Backdoors with Generative Trigger Optimization](https://mlsanity.com/paper/2511.07210/) (2025)
- [SMAGDi: Socratic Multi Agent Interaction Graph Distillation for Efficient High Accuracy Reasoning](https://mlsanity.com/paper/2511.05528/) (2025)

## Built on by (most cited first, 6 of 6)

- [SkillsVote: Lifecycle Governance of Agent Skills from Collection, Recommendation to Evolution](https://mlsanity.com/paper/2605.18401/) (2026)
- [Contextual Agentic Memory is a Memo, Not True Memory](https://mlsanity.com/paper/2604.27707/) (2026)
- [LLM Agents Are Latent Context Managers: Eliciting Self-Managed Context via State Proprioception](https://mlsanity.com/paper/2606.30005/) (2026)
- [Reviewer Scores Are Not Comparable Across Research Areas in ML Peer Review](https://mlsanity.com/paper/2607.27209/) (2026)
- [Attributing Structured-Output Gains in Function Calling: Interface Alignment versus Procedural Transfer](https://mlsanity.com/paper/2607.02595/) (2026)
- [Distilling Agentic Systems: A Roadmap across Models, Artifacts, and Harnesses](https://mlsanity.com/paper/2609.36630/) (2026)

## Abstract (arXiv)

Multi-agent systems (MAS) for structured data-science tasks externalize analytical control through workflows spanning stages, tools, shared state, verification, and repair. Distilling such workflows into a single-agent skill can reduce orchestration overhead, but it remains unclear which workflow components should cross the control boundary. We distinguish capability resources, which expand what an agent can do, from pipeline guidance, which constrains which solutions it explores. On the same causal-estimation instances, adding task-qualified source pipeline guidance to a capability-matched skill changes normalized utility by +19.6 points under method-selection accuracy but -10.3 points under numerical error. To explain this reversal, we introduce Behavior-Outcome Freedom (F), a pre-synthesis diagnostic of signed behavior-outcome rank mismatch, and formalize its candidate-conditional role through Signed Anchor-Rank Transfer. Motivated by this mechanism, we propose AdaSkill, which preserves validated capability resources, removes runtime orchestration, and conditionally inherits pipeline guidance using a calibrated rule over F. Across 16 capability-matched interventions, the native-scale Full-minus-Discard effect decreases across the continuous F scale (r = -0.80, p < 0.001), while a 15-treatment atomic sweep localizes the reversal to pipeline guidance. Across 11 datasets spanning four structured data-science task families, AdaSkill combines strong task performance with substantially lower deployment overhead.