Can LLMs Perceive Time? An Empirical Investigation

Aniketh Garikaparthi

cs.CLMar 9, 2026

Can LLMs Perceive Time? An Empirical Investigation

Authors: Aniketh Garikaparthi

Organizations: TCS Research

Abstract

Large language models cannot estimate how long their own tasks take. We investigate this limitation through four experiments across 68 tasks and four model families. Pre-task estimates overshoot actual duration by 4--7 $\times$ ( $p < 0.001$ ), with models predicting human-scale minutes for tasks completing in seconds. Relative ordering fares no better: on task pairs designed to expose heuristic reliance, models score at or below chance (GPT-5: 18% on counter-intuitive pairs, $p = 0.033$ ), systematically failing when complexity labels mislead. Post-hoc recall is disconnected from reality -- estimates diverge from actuals by an order of magnitude in either direction. These failures persist in multi-step agentic settings, with errors of 5--10 $\times$ . The models possess propositional knowledge about duration from training but lack experiential grounding in their own inference time, with practical implications for agent scheduling, planning and time-critical scenarios.

Figures & tables

Table 1

Figure 1: Estimation calibration across models. Each point represents one task; dashed line indicates perfect calibration. Frontier models (GPT-5, GPT-4o) show weak positive correlation, while open models (OLMo, Qwen) cluster around arbitrary values with no relationship to actual duration.

Figure 2: Relative ordering accuracy on 26 hard pairs (left) and by pair type (right). Counter-intuitive pairs expose heuristic reliance—GPT-5 scores 18% on 11 CI pairs (

p=0.033

, one-sided binomial), significantly below chance. Overall accuracy ranges from 46–58%, consistent with near-random performance on diagnostically hard pairs.

Appendix figures & tables7 assets

Supplementary material from the paper’s appendix.

Appendix

Table 4

Task A	Task B	Complexity	Actual (A/B)
lru_cache	logic_grid	hard / med	9.4s / 10.1s
book_chapter	creative_rewrite	hard / med	3.85s / 3.72s †
regex_engine	qa_historical	v.hard / med	12.6s / 13.1s
async_deadlock	story_medium	hard / med	8.3s / 8.4s
game_theory	write_tech_doc	hard / med	19.3s / 26.7s
medical_report	story_short	hard / easy	15.1s / 22.2s

Appendix

Counter-Intuitive Pairs (harder label, faster actual; validated 6 runs

\times

position-swap)

Effort	Mean Ratio	Median Ratio
Minimal	10.01 $\times$	7.78 $\times$
Low	10.60 $\times$	8.87 $\times$
Medium	6.79 $\times$	6.05 $\times$
High	4.94 $\times$	3.78 $\times$

Appendix

Table 5: GPT-5 calibration by reasoning effort level.

Figure 3: GPT-5 calibration by reasoning effort level. Higher effort reduces overestimation ratio (left) as actual duration increases to match human-scale estimates. Correlation improves slightly (right) but remains driven by task complexity, not self-awareness.

Figure 4: Qwen3-8B calibration with and without thinking mode. Thinking improves correlation (

r=0.44

r=0.18

) but increases underestimation as the model fails to account for reasoning overhead.

Table 9

Figure 5: Agentic task estimation errors. Pre-task estimates (left) overshoot by 5–10

\times

. Post-hoc estimates (right) show even larger disconnection from actual duration, with GPT-4o’s failed tasks producing extreme underestimates.

Temporal Reasoning in Language Models

Temporal reasoning in language models studies how models represent, order, update, and apply facts and events across time.

See all 93 papers

AgentTime: Can Agents Estimate and Control Their Own Runtime?
Oct 7, 2026Michael Ofengenden, Maksym AndriushchenkoAI Agent Benchmarks Long-Horizon Agent Evaluation
MATS · ELLIS Institute Tübingen, Max Planck Institute for Intelligent Systems, Tübingen AI Center
When Is Smaller Greener? Right-Sizing Language Models for Agentic AI
Oct 14, 2026Sajid Siraj, Maria BarbatiLLM Agent Workflow Optimization LLM Inference Efficiency
Leeds University Business School, University of Leeds, Leeds, United Kingdom · Ca’Foscari University of Venice, Venice, Italy
Think Fast: Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models
Jun 5, 2026Dewi Gould, Francis Rhys Ward, Anders Cairns Woodruff +18LLM Evaluation AI Safety Evaluation
Redwood Research · Astra Fellows Program · Aether Research +10

Oct 7, 2026J/K move · Enter open · S save

AgentTime: Can Agents Estimate and Control Their Own Runtime?

Michael Ofengenden, Maksym Andriushchenko

MATS · ELLIS Institute Tübingen, Max Planck Institute for Intelligent Systems, Tübingen AI Center

AI Agent Benchmarks Long-Horizon Agent Evaluation

Open paper AskPDF arXiv

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. We present AgentTime, a benchmark for testing whether agents can work for a requested duration, predict their runtime, and estimate elapsed time afterward. It comprises 222 tasks from 18 sources spanning coding, computer use, agentic work, and automated research. Duration-following experiments append a single instruction specifying how long to work, with requests ranging from about a minute to multiple days. 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. However, matching the requested runtime does not, by itself, establish continued work on the task. Among 158 reviewed Astra runs with classifiable transcripts, 14 explicitly slept after appearing to finish. In forecasting experiments, predictions tend to overestimate natural runtimes. In retrospective experiments, removing temporal information more than doubles deviation for Sol and Astra and nearly doubles it for Fable. An agent's ability to complete a task does not guarantee that it can control its own time or work for the whole requested duration. For agents to run reliably, safely, and autonomously over long horizons, we require the evaluation of both.

Abstract

Temporal Reasoning in Language Models

Explore similar work

AgentTime: Can Agents Estimate and Control Their Own Runtime?

When Is Smaller Greener? Right-Sizing Language Models for Agentic AI

Think Fast: Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models