Agensh Paper Scales Multi-Agent LLM System to 1,024 Agents on Organizational QA Tasks

'Agensh: Scaling Organizational Intelligence to 1,024 Agents' hit HF Papers on Sept 23, benchmarking a multi-agent LLM system across QA workloads as it scales the population of collaborating agents into the four-digit range. The paper is one of a cluster of Sept 23 arXiv drops probing how research-agent behavior changes when treated as an organization rather than a monolithic policy, and lands the same week as NVIDIA's Agora and Microsoft's ProgramDistill agent-scaling work.