FlowBank: New Framework Optimizes Multi-Agent LLM Workflows Through Adaptive Portfolio Selection
Researchers introduced FlowBank, a framework that creates a compact portfolio of reusable AI workflows optimized for different types of queries, rather than searching for a single universal workflow or generating new ones per query. The approach addresses a key trade-off in current LLM-based multi-agent systems between expensive offline computation and costly real-time generation. The method achieved 4-15% improvements over existing approaches while maintaining competitive computational costs.
FlowBank proposes a three-stage optimization framework for LLM-based multi-agent systems that balances computational efficiency with performance. The key insight is that workflows discovered during offline search often handle different subsets of queries, and many queries solvable by expensive query-level generation can already be solved by cheaper precomputed workflows. The framework operates through three components: DiverseFlow generates complementary workflow candidates by steering search toward under-covered queries; CuraFlow compresses the candidate pool into a compact, non-redundant portfolio; and a matching stage uses edge-value prediction on a query-workflow bipartite graph to route each query to the optimal workflow. Testing across five benchmarks showed FlowBank achieved the highest average scores while remaining cost-competitive, with 4.26% improvement over automated baselines and 14.92% over handcrafted ones.
What different sources said
- arXiv cs.AICenter
FlowBank: Query-Adaptive Agentic Workflows Optimization through Precompute-and-Reuse
Related
Genetic Drift, Not Selection, Drives Rapid Feather Color Evolution in Island Bird Radiation
A new study of an island bird radiation found that rapid evolution of feather coloration is driven primarily by genetic drift in small populations rather than sexual or ecological selection. The research integrated whole-genome data with detailed plumage measurements across complete species sampling to test whether signaling trait evolution correlates with speciation rates. The findings suggest that neutral demographic processes play a central role in generating phenotypic diversity during island radiations, challenging assumptions about the mechanisms driving rapid evolution.
New AI Model Improves Prediction of Therapeutic Peptide Function from Protein Sequences
Researchers developed a lightweight CNN classifier that predicts whether peptide sequences have therapeutic properties, trained on a database of 54,655 peptides across 48 functional categories. The model uses a novel negative sampling strategy to reduce false positive rates from over 60% in previous approaches to 2.1%. This advancement could accelerate drug discovery by enabling faster computational screening of peptide candidates before expensive experimental testing.
Study Shows Different Metabolic Stress Models Produce Distinct Effects on Human Neuronal Networks
Researchers tested three common in vitro metabolic stress models on human-derived neuronal networks and found each produced different patterns of neuronal activity and cell damage. The models tested were hypoxia alone, oxygen-glucose deprivation (OGD), and hypoxia combined with glutamate exposure. The findings suggest that choice of experimental model significantly affects results and that combining electrophysiological and structural analyses is important for accurately assessing metabolic stress in stroke research.