Pith. sign in

REVIEW 5 major objections 5 minor 33 references

Evaluating Temporal Plasticity in Foundation Time Series Models for Incremental Fine-tuning

T0 review · 5 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Foundation time series models keep improving as new data arrives.

desk verdict First look at temporal plasticity in time-series foundation models, but the headline claim of sustained improvement is undercut by the authors' own Rzero analysis; promising direction that needs major revision. read the letter →

arxiv 2504.14677 v1 pith:HFBRLHLX submitted 2025-04-20 cs.LG cs.AI

Jia Liu , Cheng Jinguo , Xia Fang , Zhenyuan Ma , Yuankai Wu This is my paper · ORCID
classification cs.LGcs.AI
keywords temporalplasticitytimeseriesfoundationmodelscontinuallearningincrementalfine-tuningcatastrophicforgettinglossofTime-MoEChronos
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that large pre-trained time series foundation models can be updated continuously on newly collected data without losing the ability to learn or forgetting what they already know. The authors call this capacity temporal plasticity and study it with an incremental fine-tuning pipeline that splits each dataset into ten chronological subsets. On two real-world datasets, the foundation models Time-MoE and Chronos show sustained predictive improvement under this regime, while smaller models such as DLinear, PatchTST, and iTransformer deteriorate. If this is right, practitioners could keep foundation models current by periodically fine-tuning on the newest observations instead of retraining from scratch, and the field's effort is better spent on fine-tuning strategies than on bespoke small models.

What carries the argument

The load-bearing mechanism is a continual-learning evaluation pipeline that splits each dataset chronologically into $P = 10$ subsets and compares three training regimes: zero-shot, incremental fine-tuning on each new subset, and full retraining on all data up to that point. Performance is captured by three ratios — $R^{\mathrm{zero}}_p$, $R^{\mathrm{full}}_p$, and $R^{\mathrm{fz}}_p$ — which isolate whether incremental fine-tuning beats the zero-shot model, approaches the fully retrained model, and whether full training improves over zero-shot, respectively. These ratios, rather than any single accuracy number, are what the paper uses to diagnose plasticity loss, catastrophic forgetting, and temporal adaptability.

What would settle it

Search the pretraining corpora of Time-MoE and Chronos for the Flight and CD-Bike series, or rerun the incremental fine-tuning protocol on a dataset constructed after those models were trained; if the zero-shot baselines improve because the series appeared in pretraining, the reported $R^{\mathrm{zero}}_p$ gains would shrink or invert.

Watch

Extended reading notes

Core claim

The central claim is that foundation models for time series forecasting exhibit temporal plasticity: when fine-tuned sequentially on new temporal subsets of data, Time-MoE and Chronos continue to improve predictive accuracy relative to their zero-shot baselines, whereas conventional deep learning models suffer from plasticity loss and catastrophic forgetting. The paper evaluates this with ratio metrics $R^{\mathrm{zero}}_p = \mathrm{MSE}^{\mathrm{inc}}_p / \mathrm{MSE}^{\mathrm{zero}}_p$ and $R^{\mathrm{full}}_p = \mathrm{MSE}^{\mathrm{inc}}_p / \mathrm{MSE}^{\mathrm{full}}_p$ across $P = 10$ partitions, and reports that foundation models maintain $R^{\mathrm{zero}}_p < 1$ on most subsets and stay close to full-training performance, while small models show large deteriorations, especially after abrupt distribution shifts. A further observation is that the mini, base, and large variants of the foundation models behave similarly, suggesting that beyond a certain scale the amount of pre-training data matters more than parameter count.

Load-bearing premise

The load-bearing premise is that the two evaluation datasets were not part of the foundation models' pretraining corpora, so the zero-shot baselines used to compute the improvement ratios are uncontaminated.

Editorial extensions

If this is right

  • Periodic fine-tuning on the newest data can keep a foundation forecaster current without full retraining, since incremental fine-tuning stays close to full-training performance.
  • The main research bottleneck shifts from designing domain-specific small models to designing fine-tuning and adaptation strategies for foundation models.
  • Abrupt distribution shifts still cause temporary performance spikes, so raw incremental fine-tuning is not yet a complete solution and robust adaptation under sudden shifts remains an open target.
  • Model scale beyond a threshold matters less than pre-training data scale, so future foundation models should emphasize diverse training corpora.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A testable extension is to run the same ratio-based pipeline on a dataset created after the models' pretraining cutoff; if the $R^{\mathrm{zero}}_p < 1$ pattern persists, the mitigation claim is not a contamination artifact.
  • If pre-training data scale is what drives plasticity, then a smaller architecture trained on the same massive corpus should exhibit the same temporal plasticity as Time-MoE; that scaling comparison would isolate the cause.
  • The three-ratio framework could be reused to benchmark continual-learning fixes such as replay, regularization, or parameter isolation against the raw fine-tuning baseline, since the paper only studies the raw protocol.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The manuscript studies whether time series foundation models can be incrementally fine-tuned on successively collected data without losing plasticity or suffering catastrophic forgetting. The authors propose a continual-learning pipeline with three evaluation ratios—Rzero_p (incremental fine-tuning vs. zero-shot), Rfull_p (incremental fine-tuning vs. full training), and Rfz_p (full training vs. zero-shot)—and apply it to Time-MoE, Chronos, and three smaller models on two real-world datasets, Flight and CD-Bike, with distribution shifts. The paper claims that foundation models demonstrate sustained improvement in predictive accuracy and effectively mitigate plasticity loss and catastrophic forgetting, while smaller models degrade. The contribution is framed as the first comprehensive evaluation of temporal plasticity for time series foundation models.

Significance. If the claims were fully supported, this would be a useful early result for practitioners deciding whether to keep foundation time series models current by periodic fine-tuning rather than full retraining. The paper contributes a concrete evaluation protocol, including algorithms and ratio-based metrics, and targets real distribution shifts, which is a relevant and understudied setting. It also ships the comparison across several public foundation-model checkpoints and smaller baselines, which is valuable for reproducibility. However, the headline conclusion is not supported by the paper's own reported Rzero_p analysis, and the proposed metrics do not directly measure forgetting or backward transfer. The evaluation also compares models under different prediction horizons and lacks statistical uncertainty. The study is therefore of moderate significance pending major revision.

major comments (5)
  1. [Abstract, §V-B, Eq. (12)] The abstract and conclusion claim that foundation models demonstrate 'sustained improvement in predictive accuracy' and 'effectively mitigate plasticity loss and catastrophic forgetting,' but Section V-B explicitly states: 'we do not observe a consistent decrease in Rzero_p with increasing p, indicating that our raw incremental learning approach does not progressively enhance foundation models' domain adaptation capabilities.' Rzero_p < 1 only shows that a checkpoint fine-tuned on partition p beats that same model's zero-shot checkpoint on partition p; it does not establish that later checkpoints improve over earlier ones or that adaptation improves over time. The central claim must either be supported by direct evidence of progressive improvement (e.g., a declining trend in MSE_inc_p over p, or comparisons of checkpoints across partitions) or be weakened to the claim that incremental fine-tuning can beat zero-shot on each new partition.
  2. [§III, Eq. (6); §V-B, Figs. 5–7] Catastrophic forgetting is defined in Eq. (6) as performance deterioration on old distributions after training on new data, but no experiment measures performance on earlier partitions using later checkpoints. The metrics Rzero_p, Rfull_p, and Rfz_p all compare checkpoints against zero-shot or full-training references on the currently considered partition; they cannot detect whether knowledge from partitions q < p has been forgotten. To support the forgetting-related claims, the authors should add a backward-transfer evaluation, for example by evaluating the model after step p on test sets from all earlier partitions and reporting those errors.
  3. [§IV, Eqs. (8)–(10)] The derivation of MSE as 1 + σ² + μ² assumes that the prediction Ŷ and the target Y are mutually independent, which is not true for a forecasting model whose predictions are functions of inputs correlated with the target. Even under a simple optimal conditional-mean predictor, E[ŶY] = E[Ŷ²], so the cross term does not vanish and the variance enters with the opposite sign relative to the stated formula. This derivation is not used directly in the ratio definitions, but it is presented as part of the proposed evaluation methodology and should be corrected or removed because, as written, it is mathematically incorrect.
  4. [§V, Setups; Figs. 3–7] The setup states that all models use a prediction length of 96, 'except for Chronos, which has a maximum prediction length of 64 due to its architectural constraints.' This means Chronos is evaluated on a shorter forecast horizon than all other models, and MSE is generally lower for shorter horizons. Consequently, the cross-model comparisons involving Chronos in Figures 3–7 are not on equal footing and can be misleading. Please either evaluate all models at the same horizon (e.g., 64) or explain why the different horizons do not affect the conclusions.
  5. [§V, Datasets] The paper asserts that the Flight and CD-Bike datasets were 'carefully selected' and 'verified' not to be part of foundation models' pretraining corpora, but no verification procedure is described and both datasets come from the authors' own prior publications ([30] and [31]). Since the zero-shot MSE values are the baselines for Rzero_p and Rfz_p, dataset contamination would directly inflate the reported gains. The authors should provide a concrete and auditable contamination check, such as matching against public pretraining corpus lists, documenting dataset release dates relative to model training, or reporting zero-shot error patterns that would be inconsistent with memorization.
minor comments (5)
  1. [§II-A] There is a typo in the reference to Chronos: 'Chnoros' should be 'Chronos'.
  2. [§II-B] The section heading 'Loss of Plasiticity and Catastrophic Forgetting in Deep Learning' contains a typo: 'Plasiticity' should be 'Plasticity'.
  3. [§IV, Algorithm 1] Algorithm 1 is described as an incremental fine-tuning algorithm but does not explicitly show the loop over partitions p = 0, 1, ..., P−1; the pseudocode should clarify that it is called once per partition with the previously fine-tuned model as input.
  4. [§V, Figures] All figures use 'Data Subset Index' on the x-axis; the text calls these partitions p. Use one consistent term throughout.
  5. [General] No code or configuration files are provided, which makes it difficult to reproduce the exact fine-tuning protocol, dataset splits, and model checkpoints. Releasing the evaluation code would strengthen the paper.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity by construction: the headline claim is empirically undercut by the paper's own Rzero analysis, but the metrics are independent measured MSE ratios, not fitted or self-defined outcomes.

full rationale

The paper's derivation chain is empirical rather than definitional. The central quantities Rzero_p, Rfull_p, and Rfz_p are ratios of independently measured MSE values under three protocols: incremental fine-tuning, zero-shot inference, and full training (Eqs. 12-14). None of these metrics is defined in terms of the paper's conclusion, and no parameter is fitted to a subset of data and then renamed as a prediction. The claim that foundation models 'demonstrate sustained improvement' is an interpretation of these ratios, not a quantity built into them. The paper's own Section V-B even acknowledges that 'we do not observe a consistent decrease in Rzero_p with increasing p, indicating that our raw incremental learning approach does not progressively enhance foundation models' domain adaptation capabilities.' That statement contradicts the abstract's 'sustained improvement' phrasing, but contradiction is not circularity. The two datasets, Flight [30] and CD-Bike [31], come from the authors' prior publications; however, those citations serve only as data provenance, not as a load-bearing theorem or fitted input. The evaluation is self-contained against zero-shot and full-training baselines, so the central comparison does not reduce to its own inputs by construction.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The central comparison rests on dataset choices, training hyperparameters, and the assumption that zero-shot baselines are contamination-free. Several of these are arbitrary or unreported (S), and the MSE decomposition in Section IV relies on an invalid independence assumption. No free parameters are fitted to the target result; the ratios are post-hoc summaries of measured MSE values.

free parameters (7)
  • Number of temporal partitions P = 10
    Choosing P=10 sets the granularity of distribution shifts and the number of fine-tuning steps; trends in Rzero, Rfull, and Rfz depend on this choice, which is not justified by data or theory.
  • Context length l = 96
    Input window for all models; forecasting performance is sensitive to context length, and no sensitivity analysis is reported.
  • Prediction length h = 96 for most models; 64 for Chronos
    Horizon is a modeling choice that directly affects MSE magnitude; Chronos's 64-step horizon creates a confound with all other models.
  • Learning rate = 1e-4
    Fixed across all models; fine-tuning behavior and plasticity results can change with learning rate, and no sweep is reported.
  • Number of epochs per fine-tuning step = 10
    Choice of training budget; plasticity loss and forgetting depend on how much each incremental update trains.
  • Train/validation/test split ratio = 6:2:2
    Arbitrary split of each temporal partition; different splits change test MSE and the ratios.
  • Retraining span S
    Used in Equation 5 and Algorithm 1 to define each new dataset D_p, but no value is reported, making the exact partition boundaries unreproducible.
assumptions (4)
  • domain assumption The Flight and CD-Bike datasets were not part of Time-MoE or Chronos pretraining corpora.
    Stated in Section V without a verifiable procedure; if false, zero-shot baselines are contaminated and Rzero and Rfz are not clean measures of adaptation.
  • ad hoc to paper Equations 8 through 10 treat model predictions Y_hat and targets Y as independent in the MSE decomposition, so the cross term vanishes.
    This assumption is false for forecasting because predictions are correlated with targets; the resulting claim that MSE approximately equals 1 + sigma squared + mu squared is incorrect and should not be used to interpret the metrics.
  • domain assumption Periodic fine-tuning with plain MSE loss on each new partition is a representative continual learning procedure.
    No replay, regularization, or continual-learning methods are included, so the comparison isolates naive incremental fine-tuning, not continual learning broadly.
  • domain assumption MSE of normalized outputs is comparable across models with different architectures and output distributions.
    The paper normalizes data before evaluation but does not describe the normalization or justify cross-model comparability when prediction horizons differ.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Evaluating Temporal Plasticity in Foundation Time Series Models for Incremental Fine-tuning." pith.science (2026). https://pith.science/paper/HFBRLHLX

@misc{pith2026250414677,
  author       = {Pith},
  title        = {Pith review of: Evaluating Temporal Plasticity in Foundation Time Series Models for Incremental Fine-tuning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HFBRLHLX}},
  note         = {Machine review of arXiv:2504.14677}
}
read the original abstract

Time series foundation models excel at diverse time series forecasting tasks, but their capacity for continuous improvement through incremental learning remains unexplored. We present the first comprehensive study investigating these models' temporal plasticity - their ability to progressively enhance performance through continual learning while maintaining existing capabilities. Through experiments on real-world datasets exhibiting distribution shifts, we evaluate both conventional deep learning models and foundation models using a novel continual learning framework. Our findings reveal that while traditional models struggle with performance deterioration during incremental fine-tuning, foundation models like Time-MoE and Chronos demonstrate sustained improvement in predictive accuracy. This suggests that optimizing foundation model fine-tuning strategies may be more valuable than developing domain-specific small models. Our research introduces new evaluation methodologies and insights for developing foundation time series models with robust continuous learning capabilities.

Figures

Figures reproduced from arXiv: 2504.14677 by the authors.

Figure 1
Figure 1. Challenges in Time Series Modeling. The figure illustrates model failure caused by temporal distribution shifts, such as the COVID-19 outbreak, and the divergence between historical and forecast data distributions. will face degraded performance when applied to data from a new distribution Pt(X, Y), as: E(X,Y)∼PDhis [L(fθ ∗ (X), Y)] ≪ E(X,Y)∼Pt [L(fθ ∗ (X), Y)] . (4) This raises a critical issue that warrants in-dep… view at source ↗
Figure 2
Figure 2. Data distribution shifts in the Flight and CD-Bike datasets. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Incremental Fine-Tuning Algorithm results for the Flight (top) and CD-Bike (bottom) datasets. Solid lines represent foundation models’ fine-tuning performance, dashed lines show small models’ fine-tuning results, and dotted lines indicate foundation models’ zero-shot performance. and Full Training methods, while foundation models (Time￾MoE [12], Chronos [14]) were additionally tested using Zero-Shot evaluation. All … view at source ↗
Figures from the paper (2 more)
Figure 6
Figure 6. Figure 6: Rfull p for the Flight (top) and CD-Bike (bottom) datasets. while full training effectively adapts to the shift. In contrast, the CD-Bike dataset, with its smaller distribution shifts, ex￾hibits smoother and consistently lower Rfull p values (below 3). Notably, small m…
Figure 7
Figure 7. Figure 7: Rfz p for the Flight (top) and CD-Bike (bottom) datasets. improvements through both Incremental Fine-Tuning and Full Training. In contrast, smaller models struggle to capture data characteristics through Incremental Fine-Tuning. The success of foundation models trained…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 18 canonical work pages

  1. [30]

    Msgnet: Learning multi- scale inter-series correlations for multivariate time series forecasting,

    W. Cai, Y . Liang, X. Liu, J. Feng, and Y . Wu, “Msgnet: Learning multi- scale inter-series correlations for multivariate time series forecasting,” in AAAI, vol. 38, no. 10, 2024, pp. 11 141–11 149

    work page 2024

  2. [31]

    Rethinking urban mobility prediction: A multivariate time series forecasting approach,

    J. Cheng, K. Li, Y . Liang, L. Sun, J. Yan, and Y . Wu, “Rethinking urban mobility prediction: A multivariate time series forecasting approach,” IEEE Transactions on Intelligent Transportation Systems , 2024

    work page 2024

  3. [1]

    Financial time series forecasting with deep learning: A systematic literature review: 2005–2019,

    O. B. Sezer, M. U. Gudelek, and A. M. Ozbayoglu, “Financial time series forecasting with deep learning: A systematic literature review: 2005–2019,” Applied soft computing , vol. 90, p. 106181, 2020

    2005

  4. [2]

    Unsupervised represen- tation learning for time series with temporal neighborhood coding,

    S. Tonekaboni, D. Eytan, and A. Goldenberg, “Unsupervised represen- tation learning for time series with temporal neighborhood coding,” in ICLR, 2020

    work page 2020

  5. [3]

    Diffusion convolutional recurrent neural network: Data-driven traffic forecasting,

    Y . Li, R. Yu, C. Shahabi, and Y . Liu, “Diffusion convolutional recurrent neural network: Data-driven traffic forecasting,” in ICLR, 2018

    2018

  6. [4]

    Accurate medium-range global weather forecasting with 3d neural networks,

    K. Bi, L. Xie, H. Zhang, X. Chen, X. Gu, and Q. Tian, “Accurate medium-range global weather forecasting with 3d neural networks,” Nature, vol. 619, no. 7970, pp. 533–538, 2023

    2023

  7. [5]

    Multivariate temporal convolutional network: A deep neural networks approach for multivariate time series forecasting,

    R. Wan, S. Mei, J. Wang, M. Liu, and F. Yang, “Multivariate temporal convolutional network: A deep neural networks approach for multivariate time series forecasting,” Electronics, vol. 8, no. 8, p. 876, 2019

    work page 2019

  8. [6]

    Temporal convolutional neural (tcn) network for an effective weather forecasting using time-series data from the local weather station,

    P. Hewage, A. Behera, M. Trovati, E. Pereira, M. Ghahremani, F. Palmieri, and Y . Liu, “Temporal convolutional neural (tcn) network for an effective weather forecasting using time-series data from the local weather station,” Soft Computing , vol. 24, pp. 16 453–16 482, 2020

    work page 2020

Show all 33 references
  1. [7]

    An empirical evaluation of generic convolutional and recurrent networks for sequence modeling,

    S. Bai, J. Z. Kolter, and V . Koltun, “An empirical evaluation of generic convolutional and recurrent networks for sequence modeling,” arXiv preprint arXiv:1803.01271, 2018

    2018 arXiv

  2. [8]

    Crossformer: Transformer utilizing cross- dimension dependency for multivariate time series forecasting,

    Y . Zhang and J. Yan, “Crossformer: Transformer utilizing cross- dimension dependency for multivariate time series forecasting,” in ICLR, 2023

    2023

  3. [9]

    itrans- former: Inverted transformers are effective for time series forecasting,

    Y . Liu, T. Hu, H. Zhang, H. Wu, S. Wang, L. Ma, and M. Long, “itrans- former: Inverted transformers are effective for time series forecasting,” in ICLR, 2024

    2024

  4. [10]

    Scaling-laws for large time-series models,

    T. D. Edwards, J. Alvey, J. Alsing, N. H. Nguyen, and B. D. Wandelt, “Scaling-laws for large time-series models,” arXiv preprint arXiv:2405.13867, 2024

    2024 arXiv

  5. [11]

    Unified training of universal time series forecasting transformers,

    G. Woo, C. Liu, A. Kumar, C. Xiong, S. Savarese, and D. Sahoo, “Unified training of universal time series forecasting transformers,” arXiv preprint arXiv:2402.02592 , 2024

    2024 arXiv

  6. [12]

    Time-moe: Billion-scale time series foundation models with mixture of experts,

    X. Shi, S. Wang, Y . Nie, D. Li, Z. Ye, Q. Wen, and M. Jin, “Time-moe: Billion-scale time series foundation models with mixture of experts,” arXiv preprint arXiv:2409.16040 , 2024

    2024 arXiv

  7. [13]

    A decoder-only foundation model for time-series forecasting,

    A. Das, W. Kong, R. Sen, and Y . Zhou, “A decoder-only foundation model for time-series forecasting,” in ICLR, 2024

    2024

  8. [14]

    Chronos: Learning the language of time series,

    A. F. Ansari, L. Stella, C. Turkmen, X. Zhang, P. Mercado, H. Shen, O. Shchur, S. S. Rangapuram, S. P. Arango, S. Kapoor et al., “Chronos: Learning the language of time series,” arXiv preprint arXiv:2403.07815, 2024

    2024 arXiv

  9. [15]

    Large language models are zero-shot time series forecasters,

    N. Gruver, M. Finzi, S. Qiu, and A. G. Wilson, “Large language models are zero-shot time series forecasters,” NeurIPS, vol. 36, 2024

    2024

  10. [16]

    Time series prediction method of industrial process with limited data based on transfer learning,

    X. Zhou, N. Zhai, S. Li, and H. Shi, “Time series prediction method of industrial process with limited data based on transfer learning,” IEEE Transactions on Industrial Informatics , vol. 19, no. 5, pp. 6872–6882, 2022

    2022

  11. [17]

    An adaptive continual learning method for nonstationary industrial time series prediction,

    M. Wu, X. Zhou, S. Li, and H. Shi, “An adaptive continual learning method for nonstationary industrial time series prediction,” IEEE Trans- actions on Industrial Informatics , 2024

    2024

  12. [18]

    Loss of plasticity in deep continual learning,

    S. Dohare, J. F. Hernandez-Garcia, Q. Lan, P. Rahman, A. R. Mahmood, and R. S. Sutton, “Loss of plasticity in deep continual learning,” Nature, vol. 632, no. 8026, pp. 768–774, 2024

    2024

  13. [19]

    Catastrophic interference in connec- tionist networks: The sequential learning problem,

    M. McCloskey and N. J. Cohen, “Catastrophic interference in connec- tionist networks: The sequential learning problem,” in Psychology of learning and motivation . Elsevier, 1989, vol. 24, pp. 109–165

    1989

  14. [20]

    A survey of time series foundation models: Generalizing time series representation with large language mode,

    J. Ye, W. Zhang, K. Yi, Y . Yu, Z. Li, J. Li, and F. Tsung, “A survey of time series foundation models: Generalizing time series representation with large language mode,” arXiv preprint arXiv:2405.02358 , 2024

    2024

  15. [21]

    One fits all: Power general time series analysis by pretrained lm,

    T. Zhou, P. Niu, L. Sun, R. Jin et al. , “One fits all: Power general time series analysis by pretrained lm,” NeurIPS, vol. 36, 2024

    2024

  16. [22]

    Time-LLM: Time series forecasting by reprogramming large language models,

    M. Jin, S. Wang, L. Ma, Z. Chu, J. Y . Zhang, X. Shi, P.-Y . Chen, Y . Liang, Y .-F. Li, S. Pan, and Q. Wen, “Time-LLM: Time series forecasting by reprogramming large language models,” in ICLR, 2024

    2024

  17. [23]

    Autotimes: Au- toregressive time series forecasters via large language models,

    Y . Liu, G. Qin, X. Huang, J. Wang, and M. Long, “Autotimes: Au- toregressive time series forecasters via large language models,” arXiv preprint arXiv:2402.02370, 2024

    2024 arXiv

  18. [24]

    A comprehensive survey of continual learning: theory, method and application,

    L. Wang, X. Zhang, H. Su, and J. Zhu, “A comprehensive survey of continual learning: theory, method and application,” IEEE Transactions on Pattern Analysis and Machine Intelligence , 2024

    2024

  19. [25]

    The dormant neuron phenomenon in deep reinforcement learning,

    G. Sokar, R. Agarwal, P. S. Castro, and U. Evci, “The dormant neuron phenomenon in deep reinforcement learning,” in ICML. PMLR, 2023, pp. 32 145–32 168

    2023

  20. [26]

    Maintaining plasticity in continual learning via regenerative regularization,

    S. Kumar, H. Marklund, and B. V . Roy, “Maintaining plasticity in continual learning via regenerative regularization,” 2024. [Online]. Available: https://openreview.net/forum?id=lyoOWX0e0O

    2024

  21. [27]

    Overcoming catas- trophic forgetting with hard attention to the task,

    J. Serra, D. Suris, M. Miron, and A. Karatzoglou, “Overcoming catas- trophic forgetting with hard attention to the task,” in ICML. PMLR, 2018, pp. 4548–4557

    2018

  22. [28]

    icarl: Incremental classifier and representation learning,

    S.-A. Rebuffi, A. Kolesnikov, G. Sperl, and C. H. Lampert, “icarl: Incremental classifier and representation learning,” in CVPR, 2017, pp. 2001–2010

    2017

  23. [29]

    Gradient episodic memory for continual learning,

    D. Lopez-Paz and M. Ranzato, “Gradient episodic memory for continual learning,” NeurIPS, vol. 30, 2017

    2017

  24. [32]

    Are transformers effective for time series forecasting?

    A. Zeng, M. Chen, L. Zhang, and Q. Xu, “Are transformers effective for time series forecasting?” in AAAI, vol. 37, no. 9, 2023, pp. 11 121– 11 128

    2023

  25. [33]

    A time series is worth 64 words: Long-term forecasting with transformers,

    Y . Nie, N. H. Nguyen, P. Sinthong, and J. Kalagnanam, “A time series is worth 64 words: Long-term forecasting with transformers,” in ICLR, 2018

    2018

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.