SquidAgent boosts multi-agent task speed by smart parallel work
SquidAgent: Parallelize Wisely, Coordinate Efficiently
Artificial IntelligenceComputation and LanguageMachine Learning
Summary
Running multiple AI agents in parallel to solve complex tasks often ends up slower than doing it one after another because of hidden costs like repeating work and fixing conflicting outputs. The authors identify these costs and create a rule to decide when parallelizing a task really saves time, using token count as a more reliable measure than time estimates. They build SquidAgent to smartly split work and reduce these overheads by sharing context and pre-aligning outputs. Tests show SquidAgent speeds up task completion by over twice compared to top existing systems.
What this means in practice
- •For large scale ai developers: Improve multi-agent system throughput and reduce latency when coordinating complex multi-step AI tasks with large language models.
- •For software engineers in automation: Build faster and more efficient multi-agent automation pipelines by minimizing redundant context processing and output conflicts.
Authors
Yexiong Lin, Shanshan Ye, Yu Yao, Zhen Fang, Bo Han, Tongliang Liu
Abstract
LLM-based agents solve complex multi-step tasks, but sequential execution incurs substantial latency. In principle, parallelizing work across multiple agents should yield near-linear speedups. Yet existing parallel multi-agent systems often run slower than a single-agent baseline. We attribute this gap to two hidden costs that parallel execution incurs but a serial agent avoids. First, there is a re-exploration cost: redundant effort spent by parallel workers reconstructing context that the orchestrator already possesses, such as prior decisions, that would otherwise be inherited implicitly in a serial execution. Second, there is an alignment cost: the overhead required to reconcile inconsistencies across independently generated outputs. We thus derive a principled decision criterion: a layer should be parallelized only when its critical-path cost, plus re-exploration and alignment overheads, is lower than the corresponding serial cost. While this criterion is naturally expressed in wall-clock time, we observe that LLMs are poorly calibrated when asked to estimate task duration. To address this, we instead measure cost in predicted output tokens, which we empirically find LLMs can estimate substantially more reliably than wall-clock time. Building on this token-based criterion, we propose SquidAgent. It estimates all token budgets in a single planning step, forks each worker directly from the orchestrator's session to eliminate re-exploration cost, and replaces post-hoc reconciliation with a pre-generated shared convention block that converts alignment into a bounded upfront cost. A deterministic scheduler then applies the criterion layer by layer. Empirically, SquidAgent achieves a 2.2$\times$ mean throughput improvement and a 2.6$\times$ mean wall-time speedup over Claude Code, and a 2.0$\times$ throughput improvement over the strongest multi-agent baseline.