FlowBank: New Framework Optimizes Multi-Agent LLM Workflows Through Adaptive Portfolio Selection
Researchers have proposed FlowBank, a three-stage framework that builds a compact bank of reusable AI agent workflows and routes each query to the most suitable one at inference time. Current approaches either compute a single workflow offline or generate a new one per query, both of which waste resources; FlowBank addresses this by combining precomputation with adaptive selection. The system outperforms the strongest automated and handcrafted baselines by 4.26% and 14.92% respectively across five benchmarks, while remaining cost-competitive.
FlowBank is a preprint framework submitted to arXiv that targets a recognized inefficiency in LLM-based multi-agent systems: existing workflow optimization methods either invest heavy offline compute to produce a single workflow or incur high inference costs by synthesizing a new workflow for every query. The authors' motivating analysis finds that workflows discovered during offline search tend to solve different subsets of queries, and many queries that would trigger expensive on-the-fly generation can already be handled by cheaper precomputed workflows. FlowBank addresses this with three coupled stages: DiverseFlow steers search toward under-covered queries to build a high-coverage candidate pool; CuraFlow compresses that pool into a compact, low-redundancy portfolio; and a bipartite graph-based matching module predicts the best portfolio member for each incoming query. Evaluated across five benchmarks, FlowBank achieves the highest average score among compared methods while keeping inference costs competitive. The work positions portfolio-based workflow selection as a middle path between the two dominant paradigms, suggesting that complementarity among workflows is an underexploited resource in agentic AI system design.
What's missing
The paper does not detail how performance degrades as the portfolio size is constrained further, how the framework generalizes to domains outside those benchmarked, or how sensitive results are to the quality of the initial offline search.
What different sources said
- arXiv cs.AICenter
FlowBank: Query-Adaptive Agentic Workflows Optimization through Precompute-and-Reuse
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.