arXiv：cs.AI· Manvendra Modgil·· 13 小时前
按墙钟时间校准的状态监控器在智能体节奏下不存在时刻检测区间
Bistable by Construction: Wall-Clock-Calibrated State Monitors Have No Moment-Detection Regime at Agent Cadence
AI 导读
一项预注册实验在 20 条轨迹上扫描 dt∈{0..600}s 发现，按墙钟时间校准的阈值触发器只在 dt≤60s 时静默，所有临界 dt 都落在 (1,30]s。
正文
Abstract:Runtime monitors for autonomous agents commonly threshold an accumulated internal state - a behavioural baseline, a drift statistic, or, in our prior work, a modelled affective state. We previously reported a State Saturation Trap: threshold-on-state triggers over a continuous affect engine become near-constant alarms on SWE-bench debugging agents (Modgil 2026). A post-release audit found the engine received dt=0 between actions, so its exponential decay never operated: the published trap is a pure-accumulator result. We correct the record (erratum, v2) and treat the flaw as an experiment. The key variable it exposes is whether a monitor's dynamics are calibrated in sample time (per observation, as in CUSUM) or wall-clock time (half-lives in seconds, as in affect models and EMA baselines). On fixed-rate streams these coincide; on agent streams, where inter-action time varies by orders of magnitude, they do not. A pre-registered sweep over uniform intervals (dt in {0..600}s) on 20 trajectories shows the wall-clock level trigger has two regimes: at dt<=1s a constant alarm (20/20; median 18 firings); at dt>=60s silent. Every critical dt lies in (1,30]s. A scriptable gpt-5.4-mini agent loop measures latency at median 1.53s (p90 2.33s), inside the trap regime, vindicating the empirical finding under a corrected mechanism. The structure is a property of the calibration class, not the engine: a minimal wall-clock accumulator over the raw error stream reproduces the same cliff, while a sample-time CUSUM over the identical stream is exactly dt-invariant (20/20). A rising-edge trigger with hysteresis fires 0-3 times per trajectory in every condition. We conclude that wall-clock-calibrated leaky-integrator monitors admit no regime in which they act as moment detectors on agent streams; transition detection escapes the trap at every cadence, but does not recover human intervention timing.
| Comments: | v2: erratum (Sec. 3): cadence runs were logged by a scriptable gpt-5.4-mini loop, not the Claude Code hook; the cross-instrument rho = 0.530 reflects only shared error rate; the abstract's cadence claim is scoped to the harness measured. No number changes. 11 pages, 5 figures. Sequel to arXiv:2606.04296. Code: this https URL |
| Subjects: | Software Engineering (cs.SE); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| ACM classes: | I.2.7; D.2.5 |
| Cite as: | arXiv:2606.19386 [cs.SE] |
| (or arXiv:2606.19386v2 [cs.SE] for this version) | |
| https://doi.org/10.48550/arXiv.2606.19386 arXiv-issued DOI via DataCite |
Submission history
From: Manvendra Modgil [view email]
[v1]
Mon, 15 Jun 2026 21:25:52 UTC (200 KB)
[v2]
Fri, 9 Oct 2026 06:58:07 UTC (202 KB)
来源：arXiv：cs.AI · arxiv.org