Fluency, not latency: HRI metrics offer a better benchmark for real-time AI agents
Stefania_druga · x · 2026-11-06
Stefanos Nikolaidis argues benchmarks for real-time AI agents are incomplete: latency is a property of the agent alone, but fluency is a property of the human-agent pair. A perfectly correct voice assistant that leaves users idle and loses their thread shows none of this in success rates. He proposes borrowing Guy Hoffman's HRI collaborative fluency metrics—Human Idle Time (H-IDLE), Robot Idle Time (R-IDLE), Functional Delay (F-DEL), Concurrent Activity (C-ACT), plus a subjective fluency scale—defined in terms of temporal meshing between partners, and argues they transfer directly to real-time AI systems.
More from coding & agent
- Nous Research ships Hermes Agent v0.23.0 with desktop-in-browser and unified gateway — NousResearch · 2026-11-06
- Opik Launches Generative Dashboards: Plain-Language Agent Builds Custom Charts — dl_weekly · 2026-11-06
- Aggregating all agent decisions into one prioritized 'needs me' review queue — kevinkern · 2026-11-06
- AWS WAF lets developers charge AI agents for content, paid in stablecoins via Stripe — jeff_weinstein · 2026-11-06
- Ripple-65B flops in single-GPU coding test against Qwen 27B despite SOTA claims — DZhang50 · 2026-11-06
- Essay argues ontology is AI's next frontier as reasoning outpaces our definitions of reality — _AustinCalvert_ · 2026-11-06