AI TrackerFrontier desk · AEST
Back to news
normalaitech

SwarmEvolve: Latent Agent Swarms for Evolutionary Discovery and Optimization

arXiv cs.AI · 6 Nov 2026, 04:59 AEST

SwarmEvolve introduces a framework utilizing low-cost agents in swarms for evolutionary discovery and optimization. This method significantly reduces the time needed to achieve high scores on mathematical and computational problems compared to traditional approaches.

Key facts

Read aloud

SwarmEvolve introduces a framework utilizing low-cost agents in swarms for evolutionary discovery and optimization. This method significantly reduces the time needed to achieve high scores on mathematical and computational problems compared to traditional approaches. L L M-guided evolutionary search uses A I agents to improve programs for mathematical and computational problems by iteratively refining solutions and combining high-scoring methods. We present SwarmEvolve, a framework in which hundreds of low-cost agents per round work in swarms, reason in latent space without decoding to text, and share information by passing KV caches. On twenty one problems in mathematical discovery and scientific computing, SwarmEvolve gets the highest average score and needs a median of only three to eight minutes to reach the best scores that other state-of-the-art frameworks reach in sixty minutes.

Sources in this cluster