English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Recursive Multi-Agent Systems (RecursiveMAS): Latent-Space Recursive Collaboration Between AI Agents, Explained

Forum topic · 小凯 · 2026-04-29

Summary

A detailed explainer of the paper 'Recursive Multi-Agent Systems' (arXiv:2504.20018) by researchers from Tsinghua University and UC Berkeley. The paper introduces RecursiveMAS, a framework that extends recursive reasoning from a single model to an entire multi-agent system. Its core module, RecursiveLink, connects agents into a collaborative loop where they exchange compact latent states instead of natural-language messages, enabling cross-agent latent state transfer and in-distribution latent thought generation while remaining lightweight. Training uses an inner/outer loop algorithm that achieves shared gradient-based credit assignment across recursive rounds. On 9 benchmarks spanning math, science, medicine, search, and code generation, RecursiveMAS delivers an average 8.3% accuracy gain, 1.2x-2.4x end-to-end speedup, and 34.6%-75.6% token reduction over strong baselines. The article also covers four tested collaboration modes (discussion, voting, tree search, chain-of-thought), theoretical guarantees on efficiency and gradient stability, recursive depth as a new scaling axis, and limitations such as training cost and heterogeneous-agent assumptions.

Paper Explainer: Recursive Multi-Agent Systems (RecursiveMAS)

Source: arXiv:2504.20018 [cs.AI] — authors from Tsinghua University and UC Berkeley

Key idea

Traditional multi-agent systems (MAS) behave like assembly lines: one agent researches, another outlines, a third polishes — each step runs once and hands off only final text outputs. The paper proposes RecursiveMAS, which instead lets agents collaborate recursively, like a group around a round table refining the same idea over multiple rounds.

The core metaphor: five people looking at a vase from different angles. Rather than each stating one observation and leaving, they revisit and revise their views after hearing others — a spiral of deepening understanding rather than a straight pipeline.

RecursiveLink: the core module

RecursiveLink is a lightweight module connecting agents into a collaborative loop, operating in a shared latent space:

1. Cross-agent latent state transfer — agents exchange latent states rather than natural-language text, avoiding the information loss of "compressed thoughts" summarized as words. 2. In-distribution latent thought generation — agents' latent spaces stay aligned, so each agent can interpret others' "internal language." 3. Lightweight design — minimal added computational overhead.

Think of a jazz combo improvising in a loop: the pianist reharmonizes after hearing the sax, everyone responds in real time — circular, not linear.

Inner/outer loop training

Training uses an inner/outer loop learning algorithm:

  • Inner loop: gradient flow between agents within each recursive collaboration round.
  • Outer loop: joint optimization across multiple recursion rounds.
  • The key finding is shared gradient-based credit assignment: when the system succeeds, each agent can correctly attribute its own contribution — addressing the classic credit assignment problem in multi-agent learning.

    Results

    Evaluated on 9 benchmarks covering math, science, medicine, search, and code generation:

  • Accuracy: +8.3% average (on top of strong baselines)
  • Inference speed: 1.2x–2.4x end-to-end speedup
  • Token usage: reduced by 34.6%–75.6% (a major cost saving, since agents exchange compressed latent states instead of verbose text)
  • Theoretical analysis shows RecursiveMAS is more efficient than text-based MAS and maintains stable gradients during recursive training — avoiding the information/gradient decay familiar from the telephone game.

    Recursion depth as a new scaling axis

    Beyond scaling model size or data volume, RecursiveMAS proposes recursive depth as a third axis for scaling AI systems.

    Four collaboration modes tested

  • Discussion: agents debate and refine arguments over rounds
  • Voting: recursive voting informed by richer shared context
  • Tree search: agents explore different branches while sharing findings in latent space
  • Chain-of-Thought: a reasoning chain split across agents, embedded in a recursion loop with backtracking
  • RecursiveMAS improves all four, though with varying magnitude — recursion acts as a general-purpose enhancer.

    Limitations

    1. Training complexity: inner/outer loop training costs more compute. 2. Agent heterogeneity: assumes agents can share a latent space; very different architectures (e.g., vision + language) may not fit. 3. Depth vs. latency trade-off: more recursion means deeper but slower reasoning.

    Future directions include adaptive recursive depth, hybrid recursive/linear pipelines, and human-in-the-loop recursion.

    References

  • Yang, X., Zou, J., Pan, R., et al. (2026). *Recursive Multi-Agent Systems*. arXiv:2504.20018 [cs.AI].
  • Wei, J., et al. (2022). Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. *NeurIPS*.
  • Yao, S., et al. (2023). ReAct: Synergizing Reasoning and Acting in Language Models. *ICLR*.
  • Shinn, N., et al. (2024). Reflexion: Self-Reflective Agents with Habitual Reinforcement Learning. *NeurIPS*.

Tags

#multi-agent-systems#recursive-reasoning#llm#latent-space#credit-assignment#chain-of-thought#arxiv-paper#ai-efficiency

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/177618906