Drift-to-Action Controllers: Budgeted Interventions with Online Risk Certificates

Ismail Lamaakal; Chaymae Yahyati; Khalid El Makkaoui; Ibrahim Ouahbi; Yassine Maleh

cs.LGMar 9, 2026

Drift-to-Action Controllers: Budgeted Interventions with Online Risk Certificates

Authors: Ismail Lamaakal, Chaymae Yahyati, Khalid El Makkaoui, Ibrahim Ouahbi, Yassine Maleh

Organizations: Multidisciplinary Faculty of Nador, Mohammed First University, Oujda 60000, Morocco · Laboratory LaSTI, ENSAK, Sultan Moulay Slimane University, Khouribga 54000, Morocco

Abstract

Deployed machine learning systems face distribution drift, yet most monitoring pipelines stop at alarms and leave the response underspecified under labeling, compute, and latency constraints. We introduce Drift2Act, a drift-to-action controller that treats monitoring as constrained decision-making with explicit safety. Drift2Act combines a sensing layer that maps unlabeled monitoring signals to a belief over drift types with an active risk certificate that queries a small set of delayed labels from a recent window to produce an anytime-valid upper bound $U_t(δ)$ on current risk. The certificate gates operation: if $U_t(δ) \le τ$ , the controller selects low-cost actions (e.g., recalibration or test-time adaptation); if $U_t(δ) > τ$ , it activates abstain/handoff and escalates to rollback or retraining under cooldowns. In a realistic streaming protocol with label delay and explicit intervention costs, Drift2Act achieves near-zero safety violations and fast recovery at moderate cost on WILDS Camelyon17, DomainNet, and a controlled synthetic drift stream, outperforming alarm-only monitoring, adapt-always adaptation, schedule-based retraining, selective prediction alone, and an ablation without certification. Overall, online risk certification enables reliable drift response and reframes monitoring as decision-making with safety.

Figures & tables

Figure 1: Belief evolution over drift types. Posterior $b_{t}(d)=\mathbb{P}(D_{t}=d\mid z_{1:t})$ over $\{\mathrm{none},\mathrm{covariate},\mathrm{concept},\mathrm{subgroup}\}$ across time; dashed lines mark drift events that shift belief and guide intervention choice.

Method	$C_{\mathrm{tot}}$	$V$	$T_{\mathrm{det}}$	$T_{\mathrm{rec}}$	$\min\mathrm{Acc}^{\mathrm{wg}}$
Camelyon17 stream
Alarm-only	3.2	46	22	210	0.71
Adapt-always (TTA)	58.0	3	18	74	0.79
Retrain-on-schedule	41.5	9	35	96	0.77
Selective prediction only	9.8	14	24	160	0.74
Controller (no certificate)	18.7	7	19	98	0.78

Table 1: Main streaming results under operational constraints. We report total cost

C_{\mathrm{tot}}

, safety violations

V

, detection delay

T_{\mathrm{det}}

, recovery time

T_{\mathrm{rec}}

, and minimum worst-group accuracy

\min\mathrm{Acc}^{\mathrm{wg}}

across Camelyon17, DomainNet, and SyntheticDrift-CIFAR using

\tau=0.20

\delta=0.05

, and label delay

d=50

Panel of Figure 2 — Figure 2: Certified controller performance under cost–safety trade-offs and drift. (Left) Safety–cost Pareto frontier on Camelyon17. (Right) Recovery dynamics around a drift event relative to threshold $\tau$ , with intervention times marked.

Appendix figures & tables13 assets

Supplementary material from the paper’s appendix.

Appendix

Action	Name	Operational effect	Cost (units)
$A_{0}$	No-op	Continue inference with current parameters $\theta$ and current output policy.	$0.0$
$A_{1}$	Recalibrate	Update a lightweight calibrator using $\mathcal{S}_{t}$ ; logits unchanged.	$0.2$
$A_{2}$	Test-time adapt	Apply a small number of adaptation steps on $\mathcal{B}_{t}$ ; update $\theta\leftarrow\theta^{\prime}$ .	$1.0$
$A_{3}$	Query labels	Request $k_{t}$ labels from recent inputs; add to $\mathcal{S}_{t^{\prime}}$ when they arrive.	$0.05\,k_{t}$
$A_{4}$	Retrain	Trigger retraining using $\mathcal{S}_{t}$ plus logged data; deploy updated $\theta$ .	$12.0$
$A_{5}$	Rollback	Revert to a previously validated checkpoint $\theta_{\mathrm{safe}}$ .	$1.5$

Appendix

Table 2: Action space and normalized operational costs used by the controller.

Act.	Name	Prereq./Effect	Cooldown	Cost
$A_{0}$	No-op	Keep $\theta$ .	–	$0.0$
$A_{1}$	Recalib.	Update from $\mathcal{S}_{t}$ .	–	$0.2$
$A_{2}$	TTA	Adapt on $\mathcal{B}_{t}$ .	–	$1.0$
$A_{3}$	Query	Request $k_{t}$ ; delay $d$ .	–	$0.05k_{t}$
$A_{4}$	Retrain	Retrain + deploy.	$\Delta_{\mathrm{rt}}$	$12.0$
$A_{5}$	Rollback	Restore $\theta_{\mathrm{safe}}$ .	$\Delta_{\mathrm{rb}}$	$1.5$

Appendix

Table 3: Streaming protocol action costs and prerequisites. Normalized costs and cooldowns.

Baseline	Configuration
Alarm-only	Alarm when $\\|z_{t}\\|_{2}>\theta_{\mathrm{alarm}}$ with $\theta_{\mathrm{alarm}}=2.5$ ; no actions beyond logging.
Adapt-always	Apply $A_{2}$ every step; 1 gradient step per batch; learning rate $10^{-4}$ .
Retrain-on-schedule	Trigger $A_{4}$ every $P_{\mathrm{rt}}=1200$ steps, respecting cooldown $\Delta_{\mathrm{rt}}$ .
Selective prediction only	Predict if $\max_{y}p_{\theta}(y\mid x)\geq\theta_{\mathrm{sel}}$ with $\theta_{\mathrm{sel}}=0.6$ ; otherwise abstain ( $A_{6}$ ).
Controller (no cert.)	Same as ours but remove gating: replace $U_{t}(\delta)>\tau$ branch by normal utility maximization.

Appendix

Table 4: Baseline configurations used in all streaming experiments. We report the fixed hyperparameters that fully specify each baseline policy under the common action set and costs.

Drift type $d$	$A_{1}$ (recalib.)	$A_{2}$ (TTA)	$A_{3}$ (query)	$A_{4}$ (retrain)	$A_{5}$ (rollback)	$A_{6}$ (abstain)
none	0.10	0.05	0.08	0.12	0.10	0.15
covariate	0.35	0.70	0.25	0.85	0.40	0.55
concept	0.20	0.30	0.75	1.05	0.60	0.65
subgroup	0.25	0.35	0.85	0.95	0.55	0.80

Appendix

Table 5: Gain table

G(d,a)

used by the controller. Gains are computed by offline synthetic-drift calibration as normalized windowed risk reduction (Eqs. equation 146 –equation 147 ) at horizon

H=150

with normalization

\sigma_{R}=0.10

. Rows correspond to drift types

d\in\{\mathrm{none},\mathrm{covariate},\mathrm{concept},\mathrm{subgroup}\}

and columns correspond to actions

a\in\{A_{0},\dots,A_{6}\}

Camelyon17 stream
Method variant	$C_{\mathrm{tot}}$	$V$	$T_{\mathrm{det}}$	$T_{\mathrm{rec}}$	$\min\mathrm{Acc}^{\mathrm{wg}}$	FIR
Alarm-only	$3.2(\pm 0.3)$	$46(\pm 4)$	$22(\pm 2)$	$210(\pm 18)$	$0.71(\pm 0.02)$	$0.05(\pm 0.01)$
Adapt-always (TTA)	$58.0(\pm 2.1)$	$3(\pm 1)$	$18(\pm 2)$	$74(\pm 6)$	$0.79(\pm 0.01)$	$0.82(\pm 0.03)$
Retrain-on-schedule	$41.5(\pm 1.8)$	$9(\pm 2)$	$35(\pm 3)$	$96(\pm 9)$	$0.77(\pm 0.01)$	$0.21(\pm 0.03)$
Selective prediction only	$9.8(\pm 0.6)$	$14(\pm 3)$	$24(\pm 2)$	$160(\pm 14)$	$0.74(\pm 0.02)$	$0.14(\pm 0.02)$
Controller (no certificate)	$18.7(\pm 1.1)$	$7(\pm 2)$	$19(\pm 2)$	$98(\pm 8)$	$0.78(\pm 0.01)$	$0.16(\pm 0.02)$

Appendix

Table 6: Ablations on the full streaming protocol. Metrics match the main paper: cumulative cost

C_{\mathrm{tot}}

, safety violations

V

, detection delay

T_{\mathrm{det}}

, recovery time

T_{\mathrm{rec}}

, and minimum worst-group accuracy

\min\mathrm{Acc}^{\mathrm{wg}}

. The last column reports the false-intervention rate (FIR): the fraction of time steps in which the method intervenes (

a_{t}\neq A_{0}

) while the deployed risk is already below the safety target.

Camelyon17 stream
Variant	$C_{\mathrm{tot}}$	$V$	$T_{\mathrm{det}}$	$T_{\mathrm{rec}}$	$\min\mathrm{Acc}^{\mathrm{wg}}$	FIR
Drift2Act (full)	$24.6(\pm 1.4)$	$0(\pm 0)$	$20(\pm 2)$	$62(\pm 5)$	$0.81(\pm 0.01)$	$0.10(\pm 0.02)$
No belief (single score)	$23.3(\pm 1.3)$	$4(\pm 2)$	$21(\pm 2)$	$90(\pm 8)$	$0.78(\pm 0.01)$	$0.17(\pm 0.02)$
No active labeling (fixed rate)	$24.1(\pm 1.3)$	$2(\pm 1)$	$20(\pm 2)$	$76(\pm 7)$	$0.80(\pm 0.01)$	$0.21(\pm 0.03)$
No certificate gating	$23.7(\pm 1.2)$	$6(\pm 2)$	$20(\pm 2)$	$88(\pm 8)$	$0.79(\pm 0.01)$	$0.12(\pm 0.02)$
MMD-only monitors	$25.0(\pm 1.4)$	$3(\pm 2)$	$24(\pm 3)$	$83(\pm 8)$	$0.79(\pm 0.01)$	$0.16(\pm 0.02)$

Appendix

Table 7: Internal component ablations of Drift2Act (same controller interface). Each row removes one internal module: belief (replace with single drift-score thresholds), active labeling (replace with fixed label rate), certificate gating (do not force fallback when risk may be unsafe), and monitor stack (use MMD-only rather than multi-monitor).

	Camelyon17		DomainNet		SyntheticDrift-CIFAR
Method	FIR	Heavy-FIR	FIR	Heavy-FIR	FIR	Heavy-FIR
Alarm-only	$0.05(\pm 0.01)$	$0.00(\pm 0.00)$	$0.06(\pm 0.01)$	$0.00(\pm 0.00)$	$0.04(\pm 0.01)$	$0.00(\pm 0.00)$
Adapt-always (TTA)	$0.82(\pm 0.03)$	$0.00(\pm 0.00)$	$0.85(\pm 0.03)$	$0.00(\pm 0.00)$	$0.80(\pm 0.03)$	$0.00(\pm 0.00)$
Retrain-on-schedule	$0.21(\pm 0.03)$	$0.05(\pm 0.01)$	$0.23(\pm 0.03)$	$0.06(\pm 0.01)$	$0.20(\pm 0.03)$	$0.05(\pm 0.01)$
Selective prediction only	$0.14(\pm 0.02)$	$0.00(\pm 0.00)$	$0.18(\pm 0.02)$	$0.00(\pm 0.00)$	$0.15(\pm 0.02)$	$0.00(\pm 0.00)$
Controller (no certificate)	$0.16(\pm 0.02)$	$0.02(\pm 0.01)$	$0.19(\pm 0.02)$	$0.03(\pm 0.01)$	$0.17(\pm 0.02)$	$0.02(\pm 0.01)$

Appendix

Table 8: False-intervention diagnostics. FIR measures unnecessary interventions during already-safe operation; Heavy-FIR counts unnecessary retrain/rollback events. The certified controller remains selective, while adapt-always exhibits high unnecessary intervention by design.

Delay $d$	$C_{\mathrm{tot}}$	$V$	$T_{\mathrm{det}}$	$T_{\mathrm{rec}}$
0	$23.1(\pm 1.2)$	$0(\pm 0)$	$20(\pm 2)$	$54(\pm 5)$
25	$23.8(\pm 1.3)$	$0(\pm 0)$	$20(\pm 2)$	$58(\pm 5)$
50	$24.6(\pm 1.4)$	$0(\pm 0)$	$20(\pm 2)$	$62(\pm 5)$
100	$26.2(\pm 1.6)$	$1(\pm 1)$	$21(\pm 2)$	$78(\pm 7)$

Appendix

Table 9: Sensitivity to label delay

d

(Camelyon17) Increasing delay slows recovery because verification arrives later, which delays safe exit from fallback and slows escalation decisions.

Budget $B_{\mathrm{lab}}$	$C_{\mathrm{tot}}$	$V$	$T_{\mathrm{det}}$	$T_{\mathrm{rec}}$
1000	$21.4(\pm 1.2)$	$2(\pm 1)$	$20(\pm 2)$	$86(\pm 8)$
2000	$23.0(\pm 1.3)$	$1(\pm 1)$	$20(\pm 2)$	$72(\pm 7)$
3000	$24.6(\pm 1.4)$	$0(\pm 0)$	$20(\pm 2)$	$62(\pm 5)$
5000	$26.8(\pm 1.6)$	$0(\pm 0)$	$20(\pm 2)$	$55(\pm 5)$

Appendix

Table 10: Sensitivity to labeling budget

B_{\mathrm{lab}}

(Camelyon17) Larger budgets tighten certificates faster around drift events, reducing recovery time and violations at moderate added cost.

Drift pattern	$C_{\mathrm{tot}}$	$V$	$T_{\mathrm{det}}$	$T_{\mathrm{rec}}$	FIR	$\min\mathrm{Acc}^{\mathrm{wg}}$
Sudden drift (baseline setting)	$21.2(\pm 1.2)$	$0(\pm 0)$	$15(\pm 2)$	$58(\pm 5)$	$0.09(\pm 0.02)$	$0.82(\pm 0.01)$
Gradual drift (slow ramp)	$20.6(\pm 1.2)$	$0(\pm 0)$	$19(\pm 2)$	$72(\pm 6)$	$0.10(\pm 0.02)$	$0.81(\pm 0.01)$
Gradual drift (fast ramp)	$21.5(\pm 1.2)$	$0(\pm 0)$	$16(\pm 2)$	$63(\pm 5)$	$0.09(\pm 0.02)$	$0.82(\pm 0.01)$
Recurring drift (short period)	$22.8(\pm 1.3)$	$0(\pm 0)$	$14(\pm 2)$	$61(\pm 5)$	$0.11(\pm 0.02)$	$0.81(\pm 0.01)$
Recurring drift (long period)	$21.9(\pm 1.2)$	$0(\pm 0)$	$15(\pm 2)$	$60(\pm 5)$	$0.10(\pm 0.02)$	$0.82(\pm 0.01)$

Appendix

Table 11: Gradual and recurring drift (SyntheticDrift-CIFAR) Gradual drift mainly increases recovery time, while recurring drift mainly tests whether the controller avoids repeated heavy interventions when the stream revisits regimes.

Per-label cost	$C_{\mathrm{tot}}$	$V$	$T_{\mathrm{det}}$	$T_{\mathrm{rec}}$	FIR	$\min\mathrm{Acc}^{\mathrm{wg}}$
0.01	$23.5(\pm 1.3)$	$0(\pm 0)$	$20(\pm 2)$	$58(\pm 5)$	$0.11(\pm 0.02)$	$0.81(\pm 0.01)$
0.05	$24.6(\pm 1.4)$	$0(\pm 0)$	$20(\pm 2)$	$62(\pm 5)$	$0.10(\pm 0.02)$	$0.81(\pm 0.01)$
0.10	$25.2(\pm 1.5)$	$1(\pm 1)$	$21(\pm 2)$	$70(\pm 6)$	$0.12(\pm 0.02)$	$0.80(\pm 0.01)$

Appendix

Table 12: Effect of per-label cost (Camelyon17) When labels become expensive, the controller shifts toward adaptation and conservative fallback rather than frequent audits, increasing recovery time but preserving safety.

Noise level	$C_{\mathrm{tot}}$	$V$	$T_{\mathrm{det}}$	$T_{\mathrm{rec}}$	FIR	$\min\mathrm{Acc}^{\mathrm{wg}}$
0.00	$24.6(\pm 1.4)$	$0(\pm 0)$	$20(\pm 2)$	$62(\pm 5)$	$0.10(\pm 0.02)$	$0.81(\pm 0.01)$
0.25	$25.1(\pm 1.5)$	$0(\pm 0)$	$21(\pm 2)$	$64(\pm 5)$	$0.12(\pm 0.02)$	$0.81(\pm 0.01)$
0.50	$26.0(\pm 1.6)$	$1(\pm 1)$	$22(\pm 2)$	$70(\pm 6)$	$0.15(\pm 0.03)$	$0.80(\pm 0.01)$
0.75	$27.3(\pm 1.7)$	$2(\pm 1)$	$24(\pm 3)$	$78(\pm 7)$	$0.18(\pm 0.03)$	$0.79(\pm 0.01)$

Appendix

Table 13: Robustness to monitor noise (Camelyon17) With noisier monitors, drift evidence becomes less reliable and interventions become slightly more frequent; certificate gating prevents large safety regressions by requiring verification before unsafe operation persists.

Method	$C_{\mathrm{tot}}$	$V$	$T_{\mathrm{det}}$	$T_{\mathrm{rec}}$	$\min\mathrm{Acc}^{\mathrm{wg}}$	FIR
Alarm-only	$3.0(\pm 0.3)$	$41(\pm 5)$	$20(\pm 2)$	$180(\pm 15)$	$0.73(\pm 0.01)$	$0.05(\pm 0.01)$
Adapt-always (TTA)	$56.0(\pm 2.5)$	$4(\pm 2)$	$16(\pm 2)$	$70(\pm 7)$	$0.80(\pm 0.01)$	$0.79(\pm 0.03)$
Retrain-on-schedule	$38.0(\pm 2.0)$	$9(\pm 3)$	$28(\pm 3)$	$96(\pm 9)$	$0.78(\pm 0.01)$	$0.20(\pm 0.02)$
Controller (no certificate)	$17.9(\pm 1.2)$	$6(\pm 2)$	$18(\pm 2)$	$88(\pm 9)$	$0.79(\pm 0.01)$	$0.15(\pm 0.02)$
Controller + certificate (ours)	$23.5(\pm 1.4)$	$0(\pm 0)$	$19(\pm 2)$	$60(\pm 6)$	$0.82(\pm 0.01)$	$0.11(\pm 0.02)$

Appendix

Table 14: Amazon Reviews temporal stream. The certified controller maintains low violations and strong recovery while keeping costs moderate compared to always-adapt and schedule-based retraining.

Drifting Models

Drifting models learn transport fields that move generated samples toward a target distribution, often enabling one-step generation.

See all 53 papers

SCCM : Stream Cruise Control Method for Automated Drift Detection and Adaptation
Sep 8, 2026Mohammad Abu-Shaira, Weishi ShiConcept Drift Adaptive Learning
Department of Computer Science and Engineering, University of North Texas, 1155 Union Circle #311366, Denton, 76203, Texas, USA
SafetyDrift: Predicting When AI Agents Cross the Line Before They Actually Do
Mar 28, 2026Aditya Dhodapkar, Farhaan PishoriAI Agent Safety LLM Agent Safety
Rensselaer Polytechnic Institute Department of Computer Science · Santa Clara University Department of Engineering
DriftGuard: Safety-Aware Multi-Monitor Detection and Selective Adaptation for Evolving Toxicity Moderation
Jun 27, 2026Yuting Xin, Hanyu Cai, Binqi Shen +2Concept Drift Continual Test-Time Adaptation
Department of Information and Decision Sciences University of Minnesota Minneapolis, USA · Department of Industrial Engineering and Management Sciences Northwestern University Evanston, USA · Fuqua School of Business Duke University Durham, USA +1

Sep 8, 2026J/K move · Enter open · S save

SCCM : Stream Cruise Control Method for Automated Drift Detection and Adaptation

Mohammad Abu-Shaira, Weishi Shi

Department of Computer Science and Engineering, University of North Texas, 1155 Union Circle #311366, Denton, 76203, Texas, USA

Concept Drift Adaptive Learning

Open paper AskPDF arXiv

Real-world datasets often exhibit evolving distributions, known as concept drift. Ignoring drift degrades predictive performance, while reliance on fixed hyperparameters further limits model adaptability under changing conditions. Adaptive learning addresses this challenge by continuously updating models online, allowing them to incrementally adjust and remain effective as data distributions evolve. This paper presents the Stream Cruise Control Method (SCCM), a comprehensive framework for drift detection and adaptation in online regression. SCCM enables automated adaptation through early-response, pre-update drift detection, drift magnitude quantification, KPI-window-based thresholding for local false-alarm mitigation, dynamic hyperparameter tuning, and model recalibration. SCCM also adopts an in-memory design for real-time adaptability, unlike purely reactive methods that typically activate adaptation only after performance degradation is observed. By using dynamic thresholding and remaining agnostic to data distributions, SCCM supports KPI-based monitoring across varying data streams, including high-dimensional and large-scale settings. SCCM is integrated with four online regression models and evaluated on 18 synthetic datasets covering abrupt, incremental, and alternating gradual drift, together with eight real-world datasets. The evaluation uses both R2 and MSE and compares against eight detector--adaptation baselines. Results show improved predictive performance and effective drift handling across the evaluated online regression settings.

Abstract

Drifting Models

Explore similar work

SCCM : Stream Cruise Control Method for Automated Drift Detection and Adaptation

SafetyDrift: Predicting When AI Agents Cross the Line Before They Actually Do

DriftGuard: Safety-Aware Multi-Monitor Detection and Selective Adaptation for Evolving Toxicity Moderation