PapersSelected

MAGIC Framework Builds Mixed-Granularity Agent Graphs with Dense-Reward RL

MAGIC: Mixed-Granularity Agent Graphs via Incremental Construction with Dense-Reward Reinforcement Learning

Why we picked it

The core shift is per-role granularity: instead of committing an entire organization to single agents or fixed groups, MAGIC mixes both within one graph, which cuts execution cost while keeping reusable collaboration patterns. Researchers and engineers designing multi-agent systems can use the dense-reward shaping approach to train topology generators where sparse final-answer rewards stall learning. Compared with representative topology generators evaluated in the paper, it wins across all eight benchmarks tested.

MAGIC is a reinforcement learning framework that generates mixed-granularity agent graphs for LLM-based multi-agent systems.

Body

It sequentially selects a functional role, instantiates it as either a single agent or a reusable group, and connects it to existing units.

Because final-answer rewards provide limited intermediate feedback, the method uses potential-based reward shaping with probe-based utility and structural signals while preserving cumulative task reward.

MAGIC outperforms state-of-the-art baselines across eight benchmarks and shows strong inference efficiency.

Original · arXiv cs.LG

MAGIC: Mixed-Granularity Agent Graphs via Incremental Construction with Dense-Reward Reinforcement Learning

Collaboration topology shapes both the performance and execution cost of LLM-based multi-agent systems. Because tasks differ in complexity and required capabilities, recent approaches generate task-specific collaboration graphs that specify agent participation and information flow. However, representative topology generators use either individual agents or predefined groups throughout an organization, overlooking differing collaboration needs across subtasks. Our key insight is to select granularity locally for each functional role, combining fine-grained control with reusable collaboration patterns within one organization. Learning such organizations requires exploring a combinatorial construction space with limited intermediate feedback from final-answer rewards. Therefore, we propose MAGIC, a dense-reward reinforcement learning framework for mixed-granularity graph generation. Specifically, MAGIC constructs a mixed-granularity agent graph by sequentially selecting a functional role, instantiating it as a single agent or reusable group, and connecting it to existing units. We directly optimize the construction policy using returns from trajectories sampled under the current policy and use potential-based reward shaping to provide intermediate feedback from probe-based utility and structural signals while preserving the cumulative task reward. MAGIC outperforms state-of-the-art baselines across eight benchmarks and demonstrates strong inference efficiency in our efficiency study.

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Papers" desk. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include MAGIC, multi-agent systems, reinforcement learning, open-source models.

Source attribution: curated from arXiv cs.LG (@Kairui Yang, Ziheng Yi, Xunkai Li, Minghao An, Zhanke Liu, Zekai Chen, Rong-Hua Li).