Researchers test 9 strategies to pick LLM teams, beating random selection by up to 36.1%
shangbinfeng · x · 2026-10-23
A University of Washington-led thread introduces a systematic study of model selection for multi-LLM collaboration: with 2M+ open models available, which ones should join your team?
- Proposes and evaluates 9 selection strategies: description-based diversity, item response theory and idiosyncrasy signals, up to LLM-based "recruiters" that interview candidate models
- Two candidate pools (10 specialized Qwen models, 32 community models); teams of 4 collaborate via prompt routing, multi-agent refine, DARE-TIES merging, or LLM-Blender on math, code, and truthfulness tasks
More findings in follow-up tweets.
More from Research
- 1-gram wireless mesoscope images cortex-wide activity in freely moving mice — KordingLab · 2026-10-23
- Reversible computing veteran poses open problem: universal BARCS primitives from unbiased Josephson junctions — MikePFrank · 2026-10-23
- Arc Institute uses genomic language model to map 376M prokaryotic noncoding sequences into 5.14M clusters — tedgreenwald · 2026-10-23
- The rise of AI-assisted peer review: reviewers are making it into the appendix — abursuc · 2026-10-23
- Anthropic uses Claude to formally verify security properties in 100+ open-source libraries at RISC-V assembly level — AnthropicAI · 2026-10-23
- Transformer pretraining without backprop drops from 1.76 hours to 1.6 seconds in 10 days — mayfer · 2026-10-23