AgentTime: Can Agents Estimate and Control Their Own Runtime?
We present AgentTime, a benchmark for testing whether agents can work for a requested duration, predict their runtime, and estimate elapsed time afterward.
ProofPaper ↗
Key points
- An essential control of AI agents is their ability to manage runtime.
- This ability requires a sense of time-awareness, to predict and estimate wall-clock time and to control their own actions.
- Prior work has focused on time-awareness, but duration-following and control in native agent harnesses remain unexplored.
- Accuracy on these instructions varies substantially: Fable 5.1 in Claude Code deviates from requested runtimes by a typical factor of 2.9$\times$, compared with only 1.2$\times$ for GPT-6 Astra in Codex.
Sources (1)
- [1]AgentTime: Can Agents Estimate and Control Their Own Runtime?arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 12:26 PM
We present AgentTime, a benchmark for testing whether agents can work for a requested duration, predict their runtime, and estimate elapsed time afterward.
An essential control of AI agents is their ability to manage runtime.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 2, 2026Hundreds of millions of AI agents are coming. Is there work for them?
- Oct 2, 2026Aryanutkarsh/SeeCode: Animated, explorable diagrams for Claude and other coding agents, from a short JSON spec.
- Oct 2, 2026openqodex/openqodex: Open source AI code review for Claude Code and Codex, before you push. Scanners (SAST, secrets, dependencies, lint) on the lines you changed
- Oct 1, 2026sno-ai/sno-station-skills: Sno Station Skills: 31 tested skills that turn Claude Code and Codex into one working team. Cross-vendor peer review, automatic handoff when
- Oct 1, 2026Claude-shaped science
- Sep 30, 2026Why Dwarkesh is Wrong about Computer Use + How OpenAI shipped its Jev competitor in 1 Week