// Recursive Multi-Agent Systems // Great read for the weekend. (bookmark it) Multi-agent systems...
TL;DR · AI 摘要
RecursiveMAS 提出用共享潜在空间中的递归计算替代多智能体间冗余文本通信，显著降低 token 消耗、提升推理速度与准确率。
核心要点
- 多智能体系统瓶颈在于文本消息传递引发的 token 膨胀与上下文稀释
- RecursiveMAS 将智能体协作建模为递归计算，各 agent 充当 RLM 层并传递潜表示
- 在 9 大基准测试中平均准确率提升 8.3%，token 使用减少 34.6%–75.6%，推理加速 1.2×–2.4×
结构提纲
按章节快速跳转。
思维导图
用一张图看清主题之间的关系。
查看大纲文本（无障碍 / 无 JS 友好）
- RecursiveMAS：潜空间递归多智能体
- 问题
- 文本通信瓶颈
- Token 膨胀 / 延迟 / 上下文稀释
- 方案
- 递归计算建模
- RecursiveLink 模块
- 内外环梯度共享
- 效果
- +8.3% 准确率
- 1.2×–2.4× 推理加速
- 34.6%–75.6% Token 减少
金句 / Highlights
值得收藏与分享的关键句。
Multi-agent systems often pass full text messages between agents at every step. This leads to token bloat, latency, and context dilution which all grow with the number of agents.
A multi-agent system can be treated as a recursive computation, where each agent acts like an RLM layer, iteratively passing latent representations to the next and forming a looped interaction process
Think of it as agents passing notes in their own internal language instead of rewriting everything in English each turn. Less talking, more thinking.
Across 9 benchmarks spanning math, science, medicine, search, and code generation: 8.3% average accuracy gain over baselines, 1.2×–2.4× end-to-end inference speedup, and 34.6%–75.6% reduction in token
Great read for the weekend.
(bookmark it)
Multi-agent systems often pass full text messages between agents at every step. This leads to token bloat, latency, and context dilution which all grow with the number of agents.
RecursiveMAS asks a https://t.co/wi8S2HsBXc" / X
// Recursive Multi-Agent Systems // Great read for the weekend. (bookmark it) Multi-agent systems often pass full text messages between agents at every step. This leads to token bloat, latency, and context dilution which all grow with the number of agents. RecursiveMAS asks a different question: what if agents collaborated through recursive computation in a shared latent space, instead of through text? A multi-agent system can be treated as a recursive computation, where each agent acts like an RLM layer, iteratively passing latent representations to the next and forming a looped interaction process. They introduce a RecursiveLink module that generates latent thoughts and transfers state directly between heterogeneous agents, plus an inner-outer loop learning algorithm with shared gradient-based credit assignment across the team. Think of it as agents passing notes in their own internal language instead of rewriting everything in English each turn. Less talking, more thinking. The numbers are strong. Across 9 benchmarks spanning math, science, medicine, search, and code generation: 8.3% average accuracy gain over baselines, 1.2×–2.4× end-to-end inference speedup, and 34.6%–75.6% reduction in token usage. Why does it matter? If agent-to-agent communication is the next real bottleneck (and it is), latent-space recursion is one of the cleaner ways to scale collaboration without paying a token tax for every coordination step. Paper: arxiv.org/abs/2604.25917 Learn to build effective AI agents in our academy: academy.dair.ai
