Stack agent features
Each agent's task, role, system prompt, experience, and message are read from its frozen LLM as hidden-state slots and projected into a shared width.
Agentic AI · openJiuwen Team, Huawei · 2026
From Recursion to Convolution: Parallel Self-Improvement for Multi-Agent Systems
A framework that replaces part of the sequential recursion in multi-agent self-improvement with convolutional updates: within each agent over its features, and across agents over the collaboration graph, all in latent space.
Motivation
Recursive self-improvement lets a multi-agent system feed its improvements back into the next round. Scaling it to more agents and deeper iterations runs into three problems. Each round builds on earlier outputs, so mistakes propagate. Chain-like dependencies force later updates to wait for earlier ones, which wastes parallel compute. And refining one feature or one agent in isolation ignores how roles, prompts, experience, and messages depend on each other.
ConvRSI takes a space–time view: improvement can come not only from more sequential iterations but also from more structured spatial computation within each iteration.
Method
Each agent's task, role, system prompt, experience, and message are read from its frozen LLM as hidden-state slots and projected into a shared width.
A channel-wise 1-D convolution along the feature axis lets related features refine each other. A gated readout pools them into one local message per agent.
A GCN layer updates all agents' messages at once using the adjacency matrix of the collaboration graph, so sequential, mixture, and other topologies fit the same operator.
Inner training learns prefixes from each agent's local messages independently. Outer training precomputes prefixes for all rounds through the convolutional recurrence, then optimizes on the final output.
The language models stay frozen. Only the projections, convolutions, graph layer, and latent transfer modules are trained.
Performance
75.0%
At three rounds, 3.6 points above RecursiveMAS and 5.4 above Recursive-TextMAS, with a planner–critic–solver team of small models.
+5.2 pts
With code, science, and math specialists plus a summarizer, MATH-500 accuracy rises from 60.6% to 65.8% over RecursiveMAS.
+25.6%
On MATH-500, compared with +88.6% for RecursiveMAS and +109.4% for Recursive-TextMAS. Generated tokens stay nearly flat.
| Method | MATH-500 acc. (%) | Time (s) | MBPP+ pass@1 (%) | Time (s) |
|---|---|---|---|---|
| ConvRSI | 75.0 | 1459 | 32.9 | 695 |
| RecursiveMAS (latent) | 71.4 | 2658 | 32.1 | 1249 |
| Recursive-TextMAS (text) | 69.6 | 4721 | 29.1 | 3575 |
ConvRSI also outperforms adapted ModularRSI, Hermes+, and MASS baselines on both benchmarks; MASS at Level-1 trails it by 6.3 points on MATH-500 and 4.2 on MBPP+.