Dual-Context Analog Retrieval for Time Series Forecasting

Jung Min Choi; Ngoc Son Le; Ibram Abdelmalak; Vijaya Krishna Yalavarthi; Lars Schmidt-Thieme

cs.LGOct 2, 2026

Dual-Context Analog Retrieval for Time Series Forecasting

Authors: Jung Min Choi, Ngoc Son Le, Ibram Abdelmalak, Vijaya Krishna Yalavarthi, Lars Schmidt-Thieme

Organizations: ISMLL, University of Hildesheim · VWFS Data Analytics Research Center (VWFS-DARC), Hildesheim, Germany

Abstract

Most long-term time-series forecasting models map the look-back window directly to the full horizon in a single pass. While efficient, this design does not explicitly identify which historical states are most relevant to different future segments or exploit what followed those states. Analog forecasting addresses this by retrieving past states similar to the present and using their observed continuations, but single nearest matches can be unreliable and overlapping patches may produce redundant candidates. We propose DuoTS, a Dual-Context Time Series forecasting model that uses retrieved evidence without relying on it exclusively. DuoTS first produces a base forecast with a parallel patch encoder and linear prediction head, then progressively refines it one future patch at a time. Each refinement combines two views: a current context that attends to recent tokens and captures the latest dynamics, and a detail context that provides distinct retrieved analogs together with their subsequent trajectories. Patch-wise refinement allows the model to balance these views across the forecast horizon and associate each future segment with evidence appropriate to its temporal distance from the present. Experiments on multiple real-world datasets show that DuoTS achieves state-of-the-art performance, while ablations confirm the contribution of each context. The refinement mechanism is also model-agnostic, requiring only an encoded look-back window and the future-patch position, and can therefore be integrated into existing forecasting models.

Figures & tables

Panel of Figure 1 — Figure 1: Forecast, retrieve, revise. (a) One-shot forecasting predicts the full horizon directly. (b) DuoTS revises each future patch using current and retrieved-detail evidence with learned view weights.

Figure 2: Overview of the DuoTS architecture.

Figure 3: Overview of the dual-context correction. Each future-patch query combines recent tokens with retrieved analogs and their observed successors to produce an additive correction.

		DuoTS (Ours)		SRSNet		TimeKAN		Amplifier		iTransformer		TimeMixer		PatchTST
Dataset	Horizon	MSE	MAE	MSE	MAE	MSE	MAE	MSE	MAE	MSE	MAE	MSE	MAE	MSE	MAE
ETTh1	96	0.363 $\pm 0.0008$	0.388 $\pm 0.0003$	0.366	0.394	0.370	0.396	0.373	0.399	0.386	0.405	0.372	0.401	0.377	0.397
	192	0.400 $\pm 0.0050$	0.413 $\pm 0.0033$	0.400	0.415	0.403	0.417	0.414	0.420	0.430	0.435	0.413	0.430	0.409	0.425
	336	0.401 $\pm 0.0002$	0.430 $\pm 0.0004$	0.424	0.430	0.420	0.432	0.442	0.446	0.450	0.452	0.438	0.450	0.431	0.444
	720	0.441 $\pm 0.0023$	0.457 $\pm 0.0016$	0.426	0.455	0.442	0.465	0.455	0.467	0.495	0.487	0.483	0.483	0.457	0.477
Avg		0.401	0.422	0.404	0.423	0.408	0.427	0.421	0.433	0.440	0.444	0.426	0.441	0.418	0.435

Table 1: Long-term forecasting results for DuoTS at the horizons

H\in\{96,192,336,720\}

with the horizon average of each dataset in the rows marked Avg . The best MSE and MAE per row are in bold and the second best are underlined .

Case	ETTh2		ETTm2		Weather		Electricity
Case	MSE	MAE	MSE	MAE	MSE	MAE	MSE	MAE
DuoTS	0.318	0.375	0.257	0.316	0.219	0.259	0.160	0.254
w/o second-order dynamics	0.319	0.374	0.259	0.318	0.227	0.265	0.160	0.253
w/o channel mixing	0.319	0.375	0.256	0.314	0.227	0.265	0.166	0.260
w/o intra-patch mixer	0.322	0.376	0.263	0.319	0.224	0.262	0.164	0.258
w/o patch-axis MLP	0.330	0.383	0.263	0.318	0.224	0.263	0.158	0.252

Table 2: Component ablation results averaged over the four forecasting horizons for each dataset. Each variant was tuned separately for this ablation study. The best MSE and MAE in each column are in bold and the second best are underlined . The DuoTS row exactly reproduces the corresponding Avg rows of Table 1 .

Component	ETTm2	Solar	Weather
w/o Current context	+5.25	+8.64	+13.82
w/o Detail context	+5.89	+4.20	+11.77
w/o Both	+5.37	+9.46	+18.85

Table 3: Context-component ablation on representative datasets. Entries are

\Delta

MSE (

\times 10^{-3}

), averaged over four horizons. Positive values mean that removing the indicated component hurts.

Figure 4: Full patch-wise view-weight profiles. Mean weight

\omega_{f,2}

assigned to the retrieved-detail view at future patch

f

for Weather and Solar. The current view receives

1-\omega_{f,2}

. Curves correspond to forecast horizons.

Dataset	$H$	SRSNet	TimeKAN	PatchTST	Full DuoTS
		Original $\rightarrow$ +Dual-Context Correction from DuoTS
ETTh2	96	0.271 $\rightarrow$ 0.270	0.280 $\rightarrow$ 0.267	0.274 $\rightarrow$ 0.275	0.263
	720	0.391 $\rightarrow$ 0.382	0.420 $\rightarrow$ 0.409	0.406 $\rightarrow$ 0.393	0.383
Weather	96	0.147 $\rightarrow$ 0.145	0.151 $\rightarrow$ 0.141	0.150 $\rightarrow$ 0.143	0.142
	720	0.325 $\rightarrow$ 0.314	0.317 $\rightarrow$ 0.315	0.312 $\rightarrow$ 0.315	0.312
Electricity	96	0.131 $\rightarrow$ 0.130	0.135 $\rightarrow$ 0.135	0.143 $\rightarrow$ 0.129	0.127

Table 4: Correction transfer. Within each arrow, the lower endpoint is bold. The DuoTS column is from Table 1 and is bold when it is lower than all three correction-equipped backbones. The second-lowest value among the three correction-equipped endpoints and Full DuoTS is underlined.

Appendix figures & tables8 assets

Supplementary material from the paper’s appendix.

Appendix

Benchmark	Recorded quantity	$C$	$T$	Interval	Train / val / test
ETTh1 ( Zhou et al., 2021 )	Transformer load and oil temperature	7	17,420	1 h	8,640 / 2,880 / 2,880
ETTh2 ( Zhou et al., 2021 )	Transformer load and oil temperature	7	17,420	1 h	8,640 / 2,880 / 2,880
ETTm1 ( Zhou et al., 2021 )	Transformer load and oil temperature	7	69,680	15 min	34,560 / 11,520 / 11,520
ETTm2 ( Zhou et al., 2021 )	Transformer load and oil temperature	7	69,680	15 min	34,560 / 11,520 / 11,520
Weather ( Max Planck Institute for Biogeochemistry, 2024 )	Meteorological indicators	21	52,696	10 min	36,887 / 5,270 / 10,539
Electricity ( Trindade, 2015 )	Client-level power demand	321	26,304	1 h	18,412 / 2,632 / 5,260

Appendix

Table 5: Statistics of the eight long-term forecasting benchmarks.

C

is the number of channels and

T

is the number of observations. Dataset partitions follow chronological order.

Sampled
Coordinate	Symbol	Range, cap, or fixed value
Learning rate	$\eta$	log-uniform, $[2\times 10^{-5},3\times 10^{-3}]$
Path width	$d$	$\{32,64,128,256,384,512\}$
Mixer hidden width	$d_{\mathrm{ff}}$	$\{16,32,64,128,256,512,1024\}$
Structured-path layers	$\Lambda$	$\{1,2,3\}$
Dropout		uniform, $[0,0.40]$

Appendix

Table 6: Unified hyperparameter search space for DuoTS. Derived values are computed from the sampled coordinates, and retrieval ranges are capped by the available patch grid.

Benchmark	$H$	$L$	$P$	$S$	$d$	$d_{\mathrm{ff}}$	$\Lambda$	Dropout	$\eta$	$B$
ETTh1	96	336	16	8	512	256	2	0.274	$2.585\times 10^{-4}$	32
ETTh1	192	512	32	10	384	64	2	0.200	$1.601\times 10^{-4}$	32
ETTh1	336	512	112	56	384	256	2	0.050	$9.300\times 10^{-5}$	32
ETTh1	720	336	16	8	128	128	2	0.256	$2.443\times 10^{-4}$	32
ETTh2	96	512	48	14	256	256	1	0.200	$2.324\times 10^{-3}$	256
ETTh2	192	512	24	7	32	32	1	0.300	$2.299\times 10^{-4}$	256

Appendix

Table 7: Backbone settings of the reported DuoTS models. The look-back length

L

is chosen per cell from

\{96,336,512\}

P

and

S

are the patch length and stride,

d

is the shared path width,

d_{\mathrm{ff}}

the hidden width of the mixer MLPs, and

\Lambda

the number of structured-path layers. Dropout is shared by the backbone MLPs, the channel mixer, and the correction attention.

\eta

is the Adam learning rate and

B

the batch size.

Benchmark	$H$	$B_{\mathrm{blk}}$	$K_{\mathrm{cur}}$	$K_{\mathrm{suc}}$	$K_{\mathrm{blk}}$	$K_{\mathrm{det}}$	$\rho$	$\kappa_{\mathrm{ch}}$	$d_{g}$	$\beta$	Gate init	Corr. init	Patch init
ETTh1	96	3	4	4	1	1	0	0	4	1.27	$3.03\times 10^{-3}$	0.000	0.100
ETTh1	192	4	4	3	2	1	0	0	8	2.23	$8.08\times 10^{-4}$	0.000	0.200
ETTh1	336	4	2	1	1	2	1	1	8	2.64	$2.35\times 10^{-3}$	0.000	0.200
ETTh1	720	6	3	3	2	1	0	0	4	3.01	$4.90\times 10^{-3}$	0.000	0.100
ETTh2	96	4	6	3	1	1	0	2	4	1.71	$8.45\times 10^{-4}$	0.010	0.050
ETTh2	192	4	2	3	2	1	0	4	4	2.29	$6.25\times 10^{-4}$	0.000	0.100

Appendix

Table 8: Correction and channel settings of the reported DuoTS models.

B_{\mathrm{blk}}

is the journey block width,

K_{\mathrm{cur}}

the number of newest tokens read by the current context,

K_{\mathrm{suc}}

the successor span of a retrieved analog,

K_{\mathrm{blk}}

and

K_{\mathrm{det}}

the block and patch retrieval budgets, and

\rho

the exclusion radius, which is inactive whenever

K_{\mathrm{det}}=1

. The channel columns give the neighbourhood size

\kappa_{\mathrm{ch}}

, the graph width

d_{g}

, the initial self logit

\beta

, and the initial channel gate. The final two columns give the initial correction gate and the initial patch-axis gate.

		DuoTS		RAFT		SARAF
Dataset	$H$	MSE	MAE	MSE	MAE	MSE	MAE
ETTh1	96	0.363	0.388	0.367	0.397	0.371	0.399
	192	0.400	0.413	0.411	0.427	0.404	0.420
	336	0.401	0.430	0.436	0.442	0.435	0.447
	720	0.441	0.457	0.467	0.478	0.451	0.473
	Avg	0.401	0.422	0.420	0.436	0.415	0.435

Appendix

Table 9: Reported full-horizon comparison with retrieval-augmented forecasters (lower is better). DuoTS uses the per-cell look-back selected from

\{96,336,512\}

. RAFT and SARAF use

L=720

. RAFT values are from Tables 16–17 of Han et al. (2025) , and SARAF values are from Table 4 of Zhou et al. (2026) . These are cross-paper reported values rather than a controlled rerun. Within the three displayed methods, the best value is bold and the second-best is underlined.

Dataset	$H$	Default	Random eligible	Bottom- $K$	Nearest only	Successor shuffle
ETTh1	96	0.3634	0.3666	0.3690	0.3634	0.3632
	192	0.4009	0.4027	0.4080	0.4009	0.3989
	336	0.4017	0.4017	0.4015	0.4017	0.4015
	720	0.4412	0.4416	0.4416	0.4412	0.4413
ETTh2	96	0.2632	0.2632	0.2632	0.2632	0.2634
	192	0.3125	0.3125	0.3125	0.3125	0.3125

Appendix

Table 10: Detail-retrieval perturbations that change candidate selection or successor pairing. Bold marks the lowest MSE in each row.

Dataset	$\Delta$ MSE	Steps hurt (%)	Channels hurt (%)
Solar	-9.45	24.5	0.2
Traffic	-1.25	12.9	12.4
Weather	-17.50	0.0	16.7

Appendix

Table 11: Correction decomposition averaged over four horizons.

\Delta

MSE is corrected minus base (

\times 10^{-3}

). The last columns report where the correction has larger squared error.

Dataset	Model	Params	FLOPs	FLOPs/ $C$	Epoch	Latency	Memory
		(M)	(G)	(G/ch.)	(s)	(ms/sample)	(MB)
ETTh2 ( $C=7$ , batch size 32)
	DLinear	0.10	0.001	0.0001	0.59	0.003	21
	TimeKAN	0.05	0.035	0.0050	8.54	0.410	554
	iTransformer	0.95	0.021	0.0030	1.53	0.021	47
	PatchTST	7.40	1.971	0.2816	10.96	0.348	573

Appendix

Table 12: Computational cost with

L=512

and

H=96

on one RTX 4060 Laptop GPU. ETT uses batch size 32, Electricity uses batch size 4, and Weather uses batch size 16. Smaller values are better.

Long-Term Time Series Forecasting

Long-term time series forecasting predicts distant future values from historical sequences in domains such as energy, weather, traffic, and health.

See all 114 papers

Most-Recent Anchoring with Recurrent Ordering for Time Series Forecasting
Oct 2, 2026Jung Min Choi, Ngoc Son Le, Ibram Abdelmalak +2Time Series Forecasting Long-Term Time Series Forecasting
VWFS Data Analytics Research Center (VWFS-DARC), Hildesheim, Germany
Retrieval Mechanisms Surpass Long-Context Scaling in Time Series Forecasting
May 6, 2026Rishi Ahuja, Kumar Prateek, Simranjit Singh +1Time Series Forecasting Long-Context Modeling
Department of Information Technology, Dr. B.R. Ambedkar National Institute of Technology Jalandhar, Punjab, 144008, India.
Dual-Prototype Disentanglement: A Context-Aware Enhancement Framework for Time Series Forecasting
Jan 23, 2026Haonan Yang, Jianchao Tang, Zhuo LiDisentangled Representation Learning Prototype Learning
National University of Defense Technology, Changsha, Hunan, China

Oct 2, 2026J/K move · Enter open · S save

Most-Recent Anchoring with Recurrent Ordering for Time Series Forecasting

Jung Min Choi, Ngoc Son Le, Ibram Abdelmalak, Vijaya Krishna Yalavarthi+1

VWFS Data Analytics Research Center (VWFS-DARC), Hildesheim, Germany

Time Series Forecasting Long-Term Time Series Forecasting Recurrent Neural Networks

Open paper AskPDF arXiv

Long-term forecasting models commonly process all patches in a look-back window using the same fixed stack. Older contextual patches and recent evidence therefore receive the same computational depth. Yet the information closest to the forecast and the more distant context do not contribute equally. Uniform processing leaves this distinction unexpressed in the architecture. We propose MARO, a Most-Recent Anchoring with Recurrent Ordering model that processes the look-back window from the most recent patch to the oldest. The most recent patch serves as the anchor. It initializes the latent state and conditions each subsequent step, so older patches are folded into a representation that remains centered on recent evidence. A single shared module is reused at every step, so extending the scan further into the past introduces no additional parameters. Intermediate states retained during the scan allow the forecast head to weigh short and long portions of the history separately. This expresses recency through the order of recurrent refinement. Extensive experiments across multiple real-world time series datasets show that MARO achieves state-of-the-art performance on both long-term and short-term forecasting tasks.Ablation studies examine the contribution of the main architectural components.

Abstract

Long-Term Time Series Forecasting

Explore similar work

Most-Recent Anchoring with Recurrent Ordering for Time Series Forecasting

Retrieval Mechanisms Surpass Long-Context Scaling in Time Series Forecasting

Dual-Prototype Disentanglement: A Context-Aware Enhancement Framework for Time Series Forecasting