Do Agents Stop When They Have Enough? Measuring Stopping Calibration Against Constructed Sufficiency

Tool-using LLM agents must decide when to stop gathering information and commit to an answer, but existing benchmarks cannot separate stopping from search skill because the optimal stopping point is unknown. We construct environments in which the sufficiency step k* (the step after which every fact needed to answer has been observed) is known by construction, and in which tool outputs depend only on the step index, so every model sees the same observations and only the stopping decision varies. Across 8 models and 3,526 trajectories we tested five pre-registered predictions that agents are uniformly "sufficiency-blind" over-gatherers. The universal hypothesis is not supported. Reasoning-tuned models (gpt-5-mini, gpt-5-nano, deepseek) over-gather, and gpt-5-mini shows a multi-constraint closure asymmetry (premature-stop ratio 3.43, 95% CI [2.0, 12.5]), the only prediction confirmed, and for one model only. Four smaller instruction-tuned models do the opposite: they stop before sufficiency on 32–64% of solvable tasks, mostly by abstaining, and most often on multi-constraint and open-ended questions; claude-sonnet-4.6 stops near-optimally. No model loops on unsolvable tasks (99% abstain), and budget slack barely changes slack-normalized overshoot (ρ ≈ 0.08), although raw overshoot rises when the enforced budget is loosened. In this setting the costly failure is stopping too early on complex tasks, not the over-gathering and looping that agent evaluation usually targets. The environments, harness and analysis are available from the authors on request. Preprint. Not peer reviewed.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-16
DOI
https://doi.org/10.5281/zenodo.24270235
Primary Topic
Information Retrieval and Search Behavior
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Do Agents Stop When They Have Enough? Measuring Stopping Calibration Against Constructed Sufficiency

Khaled Shaheen, Faisal Shaheen
Zenodo (CERN European Organization for Nuclear Research)
Information Retrieval and Search Behavior
preprint

Do Agents Stop When They Have Enough? Measuring Stopping Calibration Against Constructed Sufficiency

Khaled Shaheen, Faisal Shaheen
preprint en

Abstract

Tool-using LLM agents must decide when to stop gathering information and commit to an answer, but existing benchmarks cannot separate stopping from search skill because the optimal stopping point is unknown. We construct environments in which the sufficiency step k* (the step after which every fact needed to answer has been observed) is known by construction, and in which tool outputs depend only on the step index, so every model sees the same observations and only the stopping decision varies. Across 8 models and 3,526 trajectories we tested five pre-registered predictions that agents are uniformly "sufficiency-blind" over-gatherers. The universal hypothesis is not supported. Reasoning-tuned models (gpt-5-mini, gpt-5-nano, deepseek) over-gather, and gpt-5-mini shows a multi-constraint closure asymmetry (premature-stop ratio 3.43, 95% CI [2.0, 12.5]), the only prediction confirmed, and for one model only. Four smaller instruction-tuned models do the opposite: they stop before sufficiency on 32–64% of solvable tasks, mostly by abstaining, and most often on multi-constraint and open-ended questions; claude-sonnet-4.6 stops near-optimally. No model loops on unsolvable tasks (99% abstain), and budget slack barely changes slack-normalized overshoot (ρ ≈ 0.08), although raw overshoot rises when the enforced budget is loosened. In this setting the costly failure is stopping too early on complex tasks, not the over-gathering and looping that agent evaluation usually targets. The environments, harness and analysis are available from the authors on request. Preprint. Not peer reviewed.

Zenodo (CERN European Organization for Nuclear Research)
Information Retrieval and Search Behavior
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon