RG-TTA: Regime-Guided Meta-Control for Test-Time Adaptation in Streaming Time Series

Indar Kumar; Akanksha Tiwari; Sai Krishna Jasti; Ankit Hemant Lade

cs.LGMar 29, 2026

RG-TTA: Regime-Guided Meta-Control for Test-Time Adaptation in Streaming Time Series

Authors: Indar Kumar, Akanksha Tiwari, Sai Krishna Jasti, Ankit Hemant Lade

Organizations: independent researchers

Abstract

Test-time adaptation (TTA) enables neural forecasters to adapt to distribution shifts in streaming time series, but existing methods apply the same adaptation intensity regardless of the nature of the shift. We propose Regime-Guided Test-Time Adaptation (RG-TTA), a meta-controller that continuously modulates adaptation intensity based on distributional similarity to previously-seen regimes. Using an ensemble of Kolmogorov-Smirnov, Wasserstein-1, feature-distance, and variance-ratio metrics, RG-TTA computes a similarity score for each incoming batch and uses it to (i) smoothly scale the learning rate -- more aggressive for novel distributions, conservative for familiar ones -- and (ii) control gradient effort via loss-driven early stopping rather than fixed budgets, allowing the system to allocate exactly the effort each batch requires. As a supplementary mechanism, RG-TTA gates checkpoint reuse from a regime memory, loading stored specialist models only when they demonstrably outperform the current model (loss improvement >= 30%). RG-TTA is model-agnostic and strategy-composable: it wraps any forecaster exposing train/predict/save/load interfaces and enhances any gradient-based TTA method. We demonstrate three compositions -- RG-TTA, RG-EWC, and RG-DynaTTA -- and evaluate 6 update policies (3 baselines + 3 regime-guided variants) across 4 compact architectures (GRU, iTransformer, PatchTST, DLinear), 14 datasets (6 real-world multivariate benchmarks + 8 synthetic regime scenarios), and 4 forecast horizons (96, 192, 336, 720) under a streaming evaluation protocol with 3 random seeds (672 experiments total). Regime-guided policies achieve the lowest MSE in 156 of 224 seed-averaged experiments (69.6%), with RG-EWC winning 30.4% and RG-TTA winning 29.0%. Overall, RG-TTA reduces MSE by 5.7% vs TTA while running 5.5% faster; RG-EWC reduces MSE by 14.1% vs standalone EWC.

Figures & tables

Fig. 1: RG-TTA system overview. For each incoming batch

B_{t}

, the system extracts a distributional feature vector

\mathbf{r}

(Eq. 1 ) and computes a similarity score sim against stored regimes using a four-metric ensemble (Eq. 6 ). If a high-similarity checkpoint passes the loss gate (

\ell_{\text{ckpt}}<0.70\,\ell_{\text{curr}}

, Eq. 8 ), it replaces the current model. The learning rate is smoothly modulated by similarity (Eq. 7 ), and the output head is adapted with loss-driven early stopping (max 25 steps, patience 3). The adapted checkpoint and its regime features are stored in the memory

\mathcal{M}

(dashed box; 5 slots, FIFO eviction) for future reuse. See Algorithm 1 for pseudocode.

Key	Architecture	Parameters	Family
gru_small	2-layer GRU + MLP head	${\sim}71$ K	Recurrent
itransformer	iTransformer [ 18 ]	${\sim}123$ K	Attention
patchtst	PatchTST [ 19 ]	${\sim}192$ K	Attention
dlinear	DLinear [ 20 ]	${\sim}37$ K– $1.2$ M	Linear

TABLE I: Model architectures used in experiments. All policies use the same model per experiment.

#	Policy	Steps $K$	LR	Regime mem.	Key mechanism
1	TTA	20 (fixed)	$3{\times}10^{-4}$	No	Fixed-step gradient adaptation
2	EWC	15 (fixed)	$3{\times}10^{-4}$	No	Fisher-penalised adaptation
3	DynaTTA	20 (fixed)	Dynamic	No	Sigmoid LR from shift metrics
4	RG-TTA	$\leq$ 25 (early stop)	Sim-scaled	Yes	Continuous regime-guided TTA
5	RG-EWC	$\leq$ 25 (early stop)	Sim-scaled	Yes	+ EWC regularisation ( $\lambda$ =400)
6	RG-DynaTTA	$\leq$ 25 (early stop)	DynaTTA+RG	Yes	+ DynaTTA sigmoid LR

TABLE II: The 6 update policies. Policies 1–3 are baselines; 4–6 are our contributions. “Regime mem.” indicates whether the policy maintains a checkpoint library.

Policy	Wins	Win Rate
TTA	46	20.5%
EWC	13	5.8%
DynaTTA	9	4.0%
RG-TTA	65	29.0%
RG-EWC	68	30.4%
RG-DynaTTA	23	10.3%

TABLE III: Win counts across 224 seed-averaged experiments. Regime-guided policies (ours) win 156/224 (69.6%).

Baseline	RG-Variant	$\Delta$ MSE (avg)	$\Delta$ MSE (med.)	RG Wins	$p$ -value
TTA	RG-TTA	$-5.7\%$	$-5.1\%$	150/224 (67.0%)	$1.0{\times}10^{-5}$ ***
EWC	RG-EWC	$-14.1\%$	$-10.0\%$	169/224 (75.4%)	$2.4{\times}10^{-11}$ ***
DynaTTA	RG-DynaTTA	$+0.5\%$	$-3.8\%$	139/224 (62.1%)	$6.8{\times}10^{-3}$ **

TABLE IV: Head-to-head: each baseline vs its regime-guided variant.

\Delta

MSE is average relative change (negative = RG-variant is better). Wins = experiments where the RG-variant achieves lower MSE.

p

-values from one-sided Wilcoxon signed-rank tests [ 31 ] with Bonferroni correction (

\alpha=0.05/3

Fig. 2: Pair-wise MSE change from adding regime-guidance. Negative values indicate improvement. All three pairs are statistically significant (Wilcoxon, Bonferroni-corrected).

Policy	GRU-S	iTransformer	PatchTST	DLinear
TTA	15,458	21,075	717,298	17,855
EWC	16,794	24,141	813,314	18,258
DynaTTA	16,076	20,808	874,257	17,185
RG-TTA	14,092	19,202	718,642	18,782
RG-EWC	14,047	18,851	721,305	18,752
RG-DynaTTA	16,787	22,619	868,742	16,739

TABLE V: Average MSE across 14 datasets

\times

4 horizons, by policy and model. Bold = best per column. Results averaged over 3 seeds. Synthetic datasets dominate due to scale; see Table VI for real-world results.

Fig. 3: Real-world MSE by model architecture and policy. RG-TTA and RG-EWC dominate on GRU-Small and iTransformer; DLinear benefits most from RG-DynaTTA.

Fig. 4: Demšar-style critical difference diagram (Nemenyi,

\alpha=0.05

). Connected policies are not significantly different. RG-TTA and RG-EWC rank best (2.46, 2.51).

Policy	ETTh1	ETTh2	ETTm1	ETTm2	Weather	Exchange
TTA	56.15	92.54	20.88	40.53	138.99	0.01
EWC	63.80	102.21	23.40	42.89	147.67	0.01
DynaTTA	89.29	130.18	24.62	45.49	164.37	0.01
RG-TTA	53.74	78.99	18.05	37.69	145.21	0.01
RG-EWC	52.33	78.41	18.14	37.10	151.39	0.01
RG-DynaTTA	79.90	127.21	25.22	41.79	175.31	0.01

TABLE VI: Average MSE on real-world datasets (ETT, Weather, Exchange), averaged over 4 horizons, 4 models, and 3 seeds. Bold = best per column.

Policy	ETT (4)	Weather+Exch. (2)	Synth-Recurring (3)	Synth-Shock (5)
TTA	52.53	69.50	292,818	364,420
EWC	58.08	73.84	338,101	407,818
DynaTTA	72.40	82.19	366,484	429,847
RG-TTA	47.12	72.61	300,952	358,865
RG-EWC	46.50	75.70	303,245	359,054
RG-DynaTTA	68.53	87.66	357,825	432,636

TABLE VII: Average MSE by dataset category and policy, averaged across models and horizons.

Fig. 5: Per-dataset win rate (%) by policy. Real-world datasets (top) and synthetic (bottom) show similar patterns: RG-TTA and RG-EWC dominate most datasets.

Policy	GRU-S	iTransf.	PatchTST	DLinear	Overall
TTA	106.3	33.5	381.1	16.1	134.3
EWC	253.2	42.3	312.8	17.0	156.4
DynaTTA	121.9	34.7	411.8	17.2	146.4
RG-TTA	118.7	38.7	335.7	14.5	126.9
RG-EWC	286.7	55.0	360.2	19.1	180.3
RG-DynaTTA	129.0	36.9	421.2	17.0	151.0

TABLE VIII: Average total adaptation time (seconds) per experiment, by policy and model. Lower is better.

Fig. 6: MSE vs adaptation time. Left: all 14 datasets; right: real-world only. RG-TTA achieves the best trade-off (lower MSE and lower time). Square markers = regime-guided policies.

Fig. 7: MSE scaling with forecast horizon. Left: all datasets; right: real-world only. RG-TTA and RG-EWC (solid lines) outperform baselines (dashed) across all horizons.

Sweep	Value	MSE	$\Delta$ vs Default
LR scale $\gamma$ (Eq. 7 )	0.0 (no modulation)	16,068	$+17.9\%$
	0.33	14,601	$+7.2\%$
	0.67	13,624	—
	1.0	12,881	$-5.5\%$
	1.5	11,966	$-12.2\%$
Early stopping	Fixed 20 steps	15,515	$+13.9\%$

TABLE IX: Hyperparameter sensitivity (MSE averaged over 4 datasets

\times

2 horizons

\times

3 seeds). Bold = default. TTA baseline MSE = 15,515.

Similarity Method	Relative MSE vs Ensemble
Feature distance only (v0 design)	+8.3%
KS only	+2.1%
Wasserstein only	+1.8%
Variance ratio only	+5.7%
Ensemble (KS=0.3, W=0.3, F=0.2, V=0.2)	Baseline

TABLE X: Ablation: similarity metric choice. The ensemble is more robust than any single metric.

Fig. 8: Adaptation behaviour on ETTh2 (GRU-Small, H=96). Top : batch MSE across policies. Middle : regime similarity with checkpoint threshold (dashed red) and diagnostic tier boundaries. Bottom : RG-TTA’s adaptive learning rate vs TTA’s fixed rate. Higher similarity → lower LR (conservative); lower similarity → higher LR (aggressive).

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Component	Parameter	Value
Data Pipeline
	Sequence length $L$	96
	Forecast horizons $H$	96, 192, 336, 720
	Batch size (streaming)	750
	Initial training size	720
	Max batches	10

Appendix

TABLE XI: Full hyperparameter configuration across all policies.

Dataset	Source	Rows	Frequency	Season length
ETTh1, ETTh2	[ 16 ]	17,420	Hourly	24
ETTm1, ETTm2	[ 16 ]	69,680	15-min	96
Weather	[ 17 ]	52,696	10-min	144
Exchange	[ 35 ]	7,588	Daily	5
synth_* (8)	Generated	10,000–25,200	Synthetic	50

Appendix

TABLE XII: Dataset summary.

Policy	H=96	H=192	H=336	H=720
TTA	176,830	187,295	193,082	214,479
EWC	203,296	211,448	218,073	239,690
DynaTTA	216,969	227,041	232,179	252,137
RG-TTA	170,795	184,887	191,360	223,676
RG-EWC	168,993	184,932	192,007	227,024
RG-DynaTTA	212,964	224,449	234,977	252,499

Appendix

TABLE XIII: Average MSE by horizon and policy (all 14 datasets, 4 models, 3 seeds).

Test-Time Adaptation

Test-time adaptation updates or recalibrates pretrained models at inference to handle distribution shifts, often without target labels.

See all 322 papers

Training on the Future: A Delay-Aware Audit of Test-Time Adaptation for Time-Series Forecasting
Oct 8, 2026Mohamed Readh Fentazi, Mazene Ameur, Adlen KsentiniTest-Time Adaptation Time Series Forecasting
Communication Systems Department EURECOM Sophia Antipolis, France
Towards Principled Test-Time Adaptation for Time Series Forecasting
May 17, 2026Haochun Wang, Ruichen Xu, Georgios Kementzidis +3Test-Time Adaptation Time Series Forecasting
Stony Brook University
STEPS: A Temporal Smooth Error Propagation Solver on the Manifolds for Test-Time Adaptation in Time Series Forecasting
May 8, 2026Jiaqi Liu, Yifan Ouyang, Zhifei Song +2Source-Free Domain Adaptation Test-Time Adaptation
School of Artificial Intelligence and Robotics, Xiamen University Malaysia · S.M.A.R.T. NEXUS Centre of Excellence, Xiamen University Malaysia Sepang, Selangor 43900, Malaysia

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

Training on the Future: A Delay-Aware Audit of Test-Time Adaptation for Time-Series Forecasting

Mohamed Readh Fentazi, Mazene Ameur, Adlen Ksentini

Communication Systems Department EURECOM Sophia Antipolis, France

Test-Time Adaptation Time Series Forecasting Temporal Data Leakage

Open paper AskPDF arXiv

Test-time adaptation (TTA) methods for time-series forecasting update a deployed model, or a small adapter around it, from incoming ground truth. But the label of an

H

-step forecast exists only

H

steps later, and real data pipelines add further delay. We build a leakage-free harness in which the label of forecast origin

s

is released for updates only at step

s+d

with

d \ge H

, and enforce this rule inside the released code of four recent TTA methods (TAFAS, COSA, PETSA and DynaTTA), run on their own backbones and checkpoints across five benchmarks (ETTm1, ETTh2, Weather, Electricity and Traffic). As references we add two closed-form correctors: a bank of recursive least squares (RLS) filters combined by a per-coordinate median, with no tunable hyperparameters and 56 microseconds per step on the 7-channel streams, and an ELF-style linear corrector. Under causal delayed labels the picture is asymmetric. On ETTm1 every audited method genuinely adapts, yet the RLS bank still beats three of the four at a fraction of their cost; only DynaTTA beats the bank, only at the minimum causal delay, and at roughly 2,500 times the per-update cost; the ELF-style corrector beats all four. On the other four datasets, the largest statistically significant improvement any published method achieves over its own frozen checkpoint is half a percent, on all four at least one published method is significantly worse than the frozen model at the minimum causal delay, and on drift-heavy ETTh2 longer label delays make every adapter that separates from the frozen model, ours included, significantly harmful. Leaky next-step updates inflate the apparent gains of simple adapters by up to 110%, and the backbone training recipe moves frozen online error by up to a factor of 25, more than any adaptation effect we measure. We release the harness, integration patches and all cached runs.

Abstract

Test-Time Adaptation

Explore similar work

Training on the Future: A Delay-Aware Audit of Test-Time Adaptation for Time-Series Forecasting

Towards Principled Test-Time Adaptation for Time Series Forecasting

STEPS: A Temporal Smooth Error Propagation Solver on the Manifolds for Test-Time Adaptation in Time Series Forecasting