# ATLAS: A Multi-LLM Training Framework for EvoDPO with Adaptive Reference Evolution

arXiv: https://arxiv.org/abs/2602.02709 · HTML: https://mlsanity.com/paper/2602.02709/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Reinforcement learning for LLM agents, one of 16 directions in LLM agents and tool use.
- Builds on 12 of its 32 references that are in the map.
- First posted on arXiv 2 Feb 2026.

## Placed in

- LLM agents and tool use / [Reinforcement learning for LLM agents](https://mlsanity.com/agents/reinforcement-learning-for-llm-agents/)

## Builds on (most cited first, 12 of 12)

- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation](https://mlsanity.com/paper/2308.08155/) (2023)
- [AutoAgents: A Framework for Automatic Agent Generation](https://mlsanity.com/paper/2309.17288/) (2023)
- [Algorithm Discovery With LLMs: Evolutionary Search Meets Reinforcement Learning](https://mlsanity.com/paper/2504.05108/) (2025)
- [MACPO: Weak-to-Strong Alignment via Multi-Agent Contrastive Preference Optimization](https://mlsanity.com/paper/2410.07672/) (2024)
- [TPO: Aligning Large Language Models with Multi-branch & Multi-step Preference Trees](https://mlsanity.com/paper/2410.12854/) (2024)
- [Weighted-Reward Preference Optimization for Implicit Model Fusion](https://mlsanity.com/paper/2412.03187/) (2024)
- [AI Metropolis: Scaling Large Language Model-based Multi-Agent Simulation with Out-of-order Execution](https://mlsanity.com/paper/2411.03519/) (2024)
- [AgenticSciML: Collaborative Multi-Agent Systems for Emergent Discovery in Scientific Machine Learning](https://mlsanity.com/paper/2511.07262/) (2025)
- [Debate, Reflect, and Distill: Multi-Agent Feedback with Tree-Structured Preference Optimization for Efficient Language Model Enhancement](https://mlsanity.com/paper/2506.03541/) (2025)
- [Automated Code Development for PDE Solvers Using Large Language Models](https://mlsanity.com/paper/2509.25194/) (2025)
- [WarriorMath: Enhancing the Mathematical Ability of Large Language Models with a Defect-aware Framework](https://mlsanity.com/paper/2508.01245/) (2025)

## Abstract (arXiv)

Recent multi-LLM agent systems have shown promising capabilities for automated problem-solving, yet they predominantly rely on frozen agents or static fine-tuning pipelines. To address this limitation, our primary contribution is ATLAS (Adaptive Task-distributed Learning for Agentic Self-evolution), a multi-agent framework where specialized meta-agents collaboratively train and refine an active agent toward a domain-specific policy. A core challenge in iterative preference learning within these pipelines is the reliance on fixed reference models, which typically leads to overly conservative updates or training stagnation. To overcome this, the framework's algorithmic engine utilizes Evolving Direct Preference Optimization (EvoDPO). EvoDPO employs an inspection agent to perform adaptive, proxy-KL gated reference policy updates based on continuous training telemetry. We evaluate this full framework across a diverse set of challenging environments-including non-stationary contextual bandits, partial differential equations (PINNs), and combinatorial optimization tasks (TSP, Bin Packing). Through comparison against fixed-reference, adaptive-reference, and external automated-discovery baselines, our results suggest that ATLAS combines supporter-driven exploration with EvoDPO-driven stability to improve long-horizon evaluator-driven self-improvement.