Pith. sign in

REVIEW 3 major objections 6 minor 41 references

NOMADD: Numerical Optimization of Models Adapting to Data Drift

T0 review · 3 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A post-hoc procedure forecasts how a model's decision boundary drifts and improves every base family on the Drift-Resilient TabPFN benchmark.

desk verdict A careful, honest post-hoc drift-adaptation method with real promise on lightweight models; the headline cost/accuracy claims are oversold on the foundation-model instantiations, but the core idea and benchmark discipline are worth a serious referee. read the letter →

arxiv 2608.02845 v1 pith:X33PRR4Q submitted 2026-08-03 cs.LG

Swapn Shah , Keith Burghardt This is my paper · ORCID
classification cs.LG
keywords conceptdriftdatamodeladaptationlow-rankfactorizationtemporalextrapolationtabularpost-hoclearningdecisionboundary
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

Concept drift makes deployed tabular models stale before fresh labels arrive, and most remedies either require retraining or are bespoke to neural networks. The paper tries to establish that a purely post-hoc procedure can anticipate where a model's decision boundary is heading: fit the same base model on each labeled time period, measure parameter shifts against a pooled anchor, compress those shifts into a few latent trajectories with a truncated SVD, and extrapolate each with a damped, regularized linear trend. Evaluated on the 18-dataset Drift-Resilient TabPFN benchmark under the benchmark's own protocol, this extrapolation improves every base family it is applied to over its frozen version, with the largest and most consistent gains on lightweight XGBoost. If the claim holds, drift adaptation becomes a cheap add-on for already-deployed models, including trees and black-box foundation models, rather than a property that must be built into training.

What carries the argument

The central object is the parameter-delta forecast built from a low-rank temporal factorization of boundary movement. Each per-period fit is represented by a vector $\theta(g_m)$ (for trees, leaf values; for transformers, logit fields at fixed query points), and deltas $d_m$ relative to a pooled anchor are stacked into a matrix $D$. A truncated SVD $D \approx U_r \Sigma_r V_r^\top$ reduces the $M \times P$ matrix to $r$ latent temporal sequences $z_{\cdot k} = (U_r \Sigma_r)_{\cdot k}$, and each is extrapolated by the damped, regularized trend formula $\hat{z} = \ell + \frac{s}{1+\lambda}(\phi + \phi^2 + \cdots + \phi^h)$, where $\ell$ and $s$ are the level and slope of a least-squares fit, $\phi \in (0,1]$ dampens the projection, and $\lambda \geq 0$ shrinks the slope. The decoded forecast is added to the anchor with shrinkage $\alpha$, whose zero option guarantees that declining to extrapolate is always available after validation.

What would settle it

On a stream whose latent drift is generated non-linearly—say the boundary angle rotates with a quadratic schedule, or reverses direction after a phase shift—NOMADD's damped linear extrapolation should systematically overshoot or lag the held-out periods. If on such a stream the frozen anchor ($\alpha = 0$) beats the NOMADD-updated model on out-of-distribution ROC-AUC even after forward validation, the low-rank smooth-trend premise is the cause.

Watch

Extended reading notes

Core claim

NOMADD's central claim is that a sequence of ordinary per-period model fits contains enough information to forecast where the decision boundary will move next. Formally, with per-period models $g_m$ and an anchor $g_0$ pooled over all labeled periods, the deltas $d_m = \theta(g_m) - \theta(g_0)$ are stacked and truncated via SVD into $r$ latent temporal trajectories; each trajectory is extrapolated with a damped, ridge-regularized linear trend, decoded back into a predicted delta, and applied to the anchor with a validated shrinkage $\alpha \in \{0, \tfrac{1}{4}, \tfrac{1}{2}, \tfrac{3}{4}, 1\}$. The paper shows that under this procedure every base family it applies to—logistic regression, XGBoost, MLP, TabPFN, and DR-TabPFN itself—improves on average over its frozen control on the 18-dataset benchmark, and that lightweight bases achieve ROC-AUC competitive with the pretrained DR-TabPFN while running in milliseconds of inference time.

Load-bearing premise

The procedure assumes that between labeled periods the model's decision boundary moves smoothly along a small number of latent directions, so the drift can be captured by a low-rank matrix of parameter deltas and extrapolated by a linear trend; the paper concedes that abrupt changes and non-linear motion violate this, and Figure 3 shows a real dataset where the plain trend overshoots.

Editorial extensions

If this is right

  • Any model with a sequence of labeled periods can be drift-adapted post hoc by wrapping it in NOMADD, without changing its training objective, loss, or architecture.
  • Lightweight bases such as XGBoost and logistic regression come within 0.007 pooled ROC-AUC of the pretrained foundation model on the real datasets, while predicting in milliseconds on CPUs and using no GPU memory.
  • Because the frozen anchor is always among the validated candidates, NOMADD can never be forced to move a boundary that the training periods do not justify moving; it degrades gracefully to the unmodified model.
  • The same forward-validation election that chooses hyperparameters also selects among base models, so a deployment constrained to one model family can use the in-place update, while an unconstrained deployment can let NOMADD pick the best base.

Reading between the lines

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

  • A natural extension is to replace the per-factor linear trend with a learned, nonlinear predictor on the same low-rank latent trajectories; on streams where drift is curved or accelerating, such a predictor should close the gap to the oracle while preserving the method's model-agnosticity.
  • The logit-field representation already handles black-box models, so the same pipeline could be applied to text or graph models whose outputs are probabilities, provided the low-rank smoothness assumption is re-validated on those modalities; this is an inference, not a paper claim.
  • If latent drift trajectories are shared across base families on the same data, comparing the trajectories recovered from different models could serve as a diagnostic of whether drift is inherent in the data-generating process or an artifact of a particular model class.
  • The method's graceful fallback when validation fails suggests an adaptive guard against abrupt drift: trigger a retrain or a reduced update when the forward-validation error spikes, rather than extrapolating blindly; again this is an editorial suggestion, not in the paper.
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

3 major / 6 minor

Summary. The paper presents NOMADD, a post-hoc method for concept-drift adaptation. For each labeled training period it fits the base model, computes the difference from a pooled anchor in a model-specific parameter or logit space, stacks these deltas into a matrix, applies a truncated SVD, extrapolates the latent temporal trajectories with a damped ridge-regularized linear trend, and applies the predicted delta to the anchor with a validated shrinkage factor. All hyperparameters are selected by forward validation inside the training periods. The method is applied to logistic regression, XGBoost, MLP, TabPFN, and DR-TabPFN on the 18-dataset DR-TabPFN benchmark. The paper reports positive mean gains over the full benchmark for every base family, with the largest reliable gain on XGBoost, and compares accuracy and cost against DR-TabPFN.

Significance. The idea is simple and attractive: a model-agnostic, post-hoc boundary extrapolation that does not use future labels and works across tree ensembles, neural networks, and in-context learners. The evaluation is in many ways careful: it follows the DR-TabPFN benchmark protocol and metric, separates synthetic streams from real datasets in aggregates, reports significance tests, and includes ablations with disclosed caveats. If the cost and generalization questions below are resolved, the lightweight XGBoost/NOMADD instantiation would be a practically useful contribution. The paper's strengths are the honest table notes and the clear separation between the benchmark aggregate and real-data results, but the abstract and several cost claims currently outrun those results.

major comments (3)
  1. [§3.3, §5.5, Table 4] The in-context-base version of NOMADD has a transductive cost problem that is disclosed but not resolved. Section 3.3 defines θ for TabPFN/DR-TabPFN as the logit field on a fixed set of query points, so the extrapolated delta is a vector labeled by exactly those points. Table 4 reports 0.00 s prediction time because the logit field is evaluated during fitting, and Section 5.5 then states that a new batch of test points requires re-evaluating the logit fields at a cost of M context passes. If the fixed query points are the test points used in Table 2, the reported accuracy numbers are transductive and the near-zero prediction time is an artifact of fitting on the test set. If the fixed query points are training-period points only, the paper must explain how the extrapolated delta transfers to arbitrary new inputs. Please define the deployed prediction procedure, report its true per-batch cost, and either remove or qualify the order-of-magnitude inference claim for the foundation-model instantiations.
  2. [Abstract; §5.1, Table 2] The abstract's claim that the extrapolation 'improves every base family it is applied to' holds only for the full 18-dataset aggregate, and that aggregate is carried by the three synthetic streams. Table 2's real-only means are -0.004 for Logistic, -0.011 for TabPFN, and -0.006 for DR-TabPFN (p=0.010), with only XGBoost positive and significant (+0.011, p=0.005). The paper does disclose this in the table note, but the abstract and introduction state the unqualified claim. Please rephrase the headline to match the evidence, for example by claiming improvement on the benchmark aggregate and a reliable real-data gain for XGBoost, or by presenting the real-data results as the headline.
  3. [§3.1, §3.3] The method is described as post hoc and as observing 'ordinary fits' of the base learner, but the per-period fits are not ordinary for two of the five families. Parametric models are warm-started from the anchor, and XGBoost uses a bespoke procedure in which the anchor's tree structure is frozen and only leaf values are refreshed per period. Because the strongest real-data result is the XGBoost instantiation, the comparison against the frozen control may confound the constrained leaf-refit with the extrapolation itself. Please state clearly that these per-period fitting choices are part of the method, and for XGBoost add a control that performs the same per-period leaf-refresh but applies α=0, so that the extrapolation component is isolated.
minor comments (6)
  1. [§3.4, Eq. (2)] Please define ℓ_k precisely: is it the fitted level at the last training period, or the intercept of the least-squares line? The equation as written is ambiguous about the time index associated with ℓ_k.
  2. [§5.2] Please name the 'strongest instantiation' explicitly and include the aggregate score for each instantiation in the main text; relying on Supplementary Table 1 for the central comparison is inconvenient, and the current wording is unclear about which base the 0.798 or 0.795 number refers to.
  3. [§4, §5.1] Please clarify how the ±0.005 win/loss band is applied (per split or to the dataset mean) and how it relates to the Wilcoxon signed-rank test reported in Section 5.1.
  4. [§5.4, Table 3] Please clarify the relation between the 'No SVD' ablation and the rank='full' option in the validation grid; if validation can already select full rank, state what forcing no SVD changes.
  5. [Throughout] The paper repeatedly refers to supplementary tables and figures that are not included in the arXiv submission. Please make the supplement available or move the key per-dataset results into the main text.
  6. [§1, Table 2] Typos and formatting: 'comparitively' in Section 1 and inconsistent spacing or abutting values in Table 2 should be corrected.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the forecast is a genuine out-of-sample extrapolation, and the in-context cost caveat is disclosed rather than hidden.

full rationale

The derivation chain is self-contained and not circular. NOMADD's core construction (Sections 3.3 and 3.4) fits per-period models, forms boundary deltas d_m = θ(g_m) − θ(g_0), compresses them with a truncated SVD, and extrapolates each latent trajectory with Eq. (2). The forecast ẑ_{M−1+h,k} is a function of training-period data only: no future labels enter, and the forward-validation protocol in Section 3.5 selects r, φ, λ, and α on held-out training positions. The reported gains are measured against each base's frozen control (α=0), which is a genuine external benchmark: because α=0 is always available, the method can decline to move the boundary, so an average positive gain is an empirical result rather than an identity. The comparison with DR-TabPFN uses the benchmark's own protocol, metric, and datasets, and the TabPFN/DR-TabPFN checkpoints are external artifacts, not outputs of NOMADD. The only mild issue is cost accounting for the in-context instantiations: Table 4 reports 0.00s prediction because logit fields are evaluated during fitting, and Section 5.5 explicitly discloses this ('Prediction time for the in-context instantiations is near zero because their logit fields are evaluated during fitting, not because prediction is free'). This is a transparency caveat about where compute is charged, not a circular derivation: the extrapolation itself is not fitted to test labels, and the same section reports the 101–141s fitting costs. The use of unlabeled query-point features for in-context models is transductive but disclosed, and it does not make the forecast equivalent to its inputs. There are no load-bearing self-citations; the prior work cited (DR-TabPFN, TabPFN, LoRA) is external and independently published. Section 6.1 concedes that abrupt changes and non-linear motion violate the method's assumptions, which further confirms that the predictions are not forced by construction.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

NOMADD's central claim rests on four domain assumptions about drift (low-rank, smooth, gradual, fixed features) and on the reliability of per-period fits. The free hyperparameters (r, phi, lambda, alpha) are selected by forward validation on training periods, so they are not fit to the test labels. No new physical entities are introduced.

free parameters (4)
  • rank r (SVD truncation) = selected from {1, 2, full} by forward validation
    Controls the number of latent temporal trajectories; chosen on the last V=3 training positions (Section 3.5).
  • damping phi = selected from {0.4, 0.7, 1.0} by forward validation
    Damps the extrapolated linear trend; see Section 3.4 and Equation (2).
  • ridge shrinkage lambda = selected from {0, 1} by forward validation
    Shrinks the slope of the trend; chosen by forward validation (Section 3.5).
  • shrinkage alpha = selected from {0, 0.25, 0.5, 0.75, 1} by forward validation
    Scales the forecast delta applied to the anchor; alpha=0 deploys the frozen anchor (Section 3.4, Equation (3)).
assumptions (5)
  • domain assumption Boundary movement is low-rank, i.e., per-period deltas lie near a small number of directions.
    The method compresses D with a truncated SVD (Section 3.4); rank is validated per dataset but the low-rank structure is assumed.
  • domain assumption Latent trajectories are smooth enough for a linear trend forecast.
    Equation (2) fits a damped linear trend to each latent factor; Section 6.1 acknowledges non-linear changes and Figure 3 shows overshoot on real data.
  • domain assumption Drift is gradual; abrupt changes cannot be anticipated.
    Section 6.1 states that any trend extrapolation method cannot anticipate abrupt changes.
  • domain assumption The feature set is fixed and sufficient.
    Section 6.1 assumes features are sufficient and fixed; the author notes new features may be needed to capture rapid changes.
  • domain assumption Per-period fits are reliable enough to carry the drift signal.
    Section 6.1 notes that periods of a few dozen rows put a noise floor under the delta estimates.

how reviews work

0 comments
Cite this review

Pith. "Pith review of NOMADD: Numerical Optimization of Models Adapting to Data Drift." pith.science (2026). https://pith.science/paper/X33PRR4Q

@misc{pith2026260802845,
  author       = {Pith},
  title        = {Pith review of: NOMADD: Numerical Optimization of Models Adapting to Data Drift},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/X33PRR4Q}},
  note         = {Machine review of arXiv:2608.02845}
}
read the original abstract

Tabular model performance degrades when feature distributions change over time or the relationship between features and outcome variables change over time, known as data drift and concept drift, respectively. These issues are challenging to mitigate in real time because labeled data may not be immediately available, or re-training a model could be impractical. While tools exist to reduce drift, they are typically bespoke to neural network architectures and adapt how models are trained. In this paper, we offer an alternative post-hoc method to reduce concept drift, which is applicable to a variety of models, from trees to neural networks to tabular foundation models. This new tool is especially useful when constraints, such as high model accuracy, bounded inference time, or model size requires users to choose between different models for their specific use-cases. Our algorithm fits the base model separately on each labeled training period, measures how its parameters evolve against a single anchor model pooled over all of those periods, compresses those changes with a low-rank factorization, and extrapolates each latent factor forward with a damped, regularized forecast. On the 18-dataset Drift-Resilient TabPFN benchmark, evaluated under that benchmark's own protocol and metric, the extrapolation improves every base family it is applied to, and achieves performance competitive with the state-of-the-art Drift-Resilient TabPFN with seconds of training. In contrast, Drift-Resilient TabPFN requires pre-training on millions of synthetic datasets over approximately 1,300 GPU-hours, and is orders of magnitude slower in inference (depending on the model). In the discussion, we explore the promise and challenges of extending this tool to other modalities.

Figures

Figures reproduced from arXiv: 2608.02845 by the authors.

Figure 1
Figure 1. NOMADD’s parameter evolution component. Boundary movement across labeled training periods is mea￾sured against a pooled anchor, compressed to a few latent temporal factors, extrapolated with a damped trend, and ap￾plied with a validated shrinkage α. All hyperparameters are selected by forward validation inside the training periods, and no future labels are used. 3.5 Configuration selection The rank r ∈ {1, 2, full},… view at source ↗
Figure 2
Figure 2. Decision-boundary evolution on 2-Moons for three base families across two training and two out-of-distribution test [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Boundary parameter trajectories over time for (a) [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Time and memory cost of NOMADD and DR-TabPFN. (a) Prediction time and (b) model training time (or in-context [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 28 canonical work pages

  1. [1]

    Drift-Resilient

    Helli, Kai and Schnurr, David and Hollmann, Noah and M. Drift-Resilient. Advances in Neural Information Processing Systems , volume =

    work page

  2. [2]

    , author=

    Gaussian mixture models. , author=. Encyclopedia of biometrics , volume=. 2009 , publisher=

    work page 2009

  3. [3]

    2008 Eighth IEEE International Conference on Data Mining , pages=

    Paired learners for concept drift , author=. 2008 Eighth IEEE International Conference on Data Mining , pages=. 2008 , organization=

    work page 2008

  4. [4]

    Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V

    In-context learning of evolving data streams with tabular foundational models , author=. Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 1 , pages=

    work page

  5. [5]

    arXiv preprint arXiv:2401.08406 , year=

    RAG vs fine-tuning: pipelines, tradeoffs, and a case study on agriculture , author=. arXiv preprint arXiv:2401.08406 , year=

    arXiv

  6. [6]

    2024 12th International Conference on Affective Computing and Intelligent Interaction (ACII) , pages=

    The strong pull of prior knowledge in large language models and its impact on emotion recognition , author=. 2024 12th International Conference on Affective Computing and Intelligent Interaction (ACII) , pages=. 2024 , organization=

    work page 2024

  7. [7]

    Procedia computer science , volume=

    A Survey on RAG with LLMs , author=. Procedia computer science , volume=. 2024 , publisher=

    2024

  8. [8]

    Proceedings of the sixth ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

    Mining high-speed data streams , author=. Proceedings of the sixth ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

Show all 41 references
  1. [9]

    Concept Drift Adaptation by Exploiting Historical Knowledge , year=

    Sun, Yu and Tang, Ke and Zhu, Zexuan and Yao, Xin , journal=. Concept Drift Adaptation by Exploiting Historical Knowledge , year=

  2. [10]

    Identifying Shifts in Collective Attention to Topics on Social Media

    He, Yuzi and Rao, Ashwin and Burghardt, Keith and Lerman, Kristina. Identifying Shifts in Collective Attention to Topics on Social Media. Social, Cultural, and Behavioral Modeling. 2021

    2021

  3. [11]

    2021 International Conference on Data Mining Workshops (ICDMW) , pages=

    Drift lens: Real-time unsupervised concept drift detection by evaluating per-label embedding distributions , author=. 2021 International Conference on Data Mining Workshops (ICDMW) , pages=. 2021 , organization=

    2021

  4. [12]

    IEEE transactions on knowledge and data engineering , volume=

    Learning under concept drift: A review , author=. IEEE transactions on knowledge and data engineering , volume=. 2018 , publisher=

    2018

  5. [13]

    Proceedings of the IEEE international conference on computer vision , pages=

    Revisiting unreasonable effectiveness of data in deep learning era , author=. Proceedings of the IEEE international conference on computer vision , pages=

  6. [14]

    2018 , publisher=

    Statistical properties of population stability index , author=. 2018 , publisher=

    2018

  7. [15]

    arXiv preprint arXiv:1912.13213 , year=

    A modern introduction to online learning , author=. arXiv preprint arXiv:1912.13213 , year=

    1912 arXiv

  8. [16]

    2022 , school=

    Analysing and improving the security of contactless payment cards , author=. 2022 , school=

    2022

  9. [17]

    Credit card fraud detection and concept-drift adaptation with delayed supervised information , year=

    Dal Pozzolo, Andrea and Boracchi, Giacomo and Caelen, Olivier and Alippi, Cesare and Bontempi, Gianluca , booktitle=. Credit card fraud detection and concept-drift adaptation with delayed supervised information , year=

  10. [18]

    arXiv preprint arXiv:2012.03749 , year=

    Explainable ai for interpretable credit scoring , author=. arXiv preprint arXiv:2012.03749 , year=

    2012 arXiv

  11. [19]

    New Media & Society , volume=

    Memes as games: The evolution of a digital discourse online , author=. New Media & Society , volume=. 2018 , publisher=

    2018

  12. [20]

    Social Media+ Society , volume=

    You can (not) say what you want: Using algospeak to contest and evade algorithmic content moderation on TikTok , author=. Social Media+ Society , volume=. 2023 , publisher=

    2023

  13. [21]

    Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining , pages=

    Xgboost: A scalable tree boosting system , author=. Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining , pages=

  14. [22]

    Advances in Neural Information Processing Systems , volume=

    Label delay in online continual learning , author=. Advances in Neural Information Processing Systems , volume=

  15. [23]

    Proceedings of Machine Learning and Systems , volume=

    Matchmaker: Data drift mitigation in machine learning for large-scale systems , author=. Proceedings of Machine Learning and Systems , volume=

  16. [24]

    Computers & operations research , volume=

    Neural network credit scoring models , author=. Computers & operations research , volume=. 2000 , publisher=

    2000

  17. [25]

    Journal of Accounting Research , volume=

    Detecting accounting fraud in publicly traded US firms using a machine learning approach , author=. Journal of Accounting Research , volume=. 2020 , publisher=

    2020

  18. [26]

    Statistical applications in genetics and molecular biology , volume=

    A shrinkage approach to large-scale covariance matrix estimation and implications for functional genomics , author=. Statistical applications in genetics and molecular biology , volume=. 2005 , publisher=

    2005

  19. [27]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Adversarial discriminative domain adaptation , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  20. [28]

    , author=

    Lora: Low-rank adaptation of large language models. , author=. Iclr , volume=

  21. [29]

    Proceedings of the 2015 internet measurement conference , pages=

    Peeking beneath the hood of uber , author=. Proceedings of the 2015 internet measurement conference , pages=

    2015

  22. [30]

    International Conference on Learning Representations , year =

    Hollmann, Noah and M. International Conference on Learning Representations , year =

  23. [31]

    ACM Computing Surveys , volume =

    A Survey on Concept Drift Adaptation , author =. ACM Computing Surveys , volume =

  24. [32]

    ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages=

    Forward-backward splitting for optimal transport based problems , author=. ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages=. 2020 , organization=

    2020

  25. [33]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Learning future classifiers without additional data , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  26. [34]

    IEEE Transactions on Knowledge and Data Engineering , volume=

    Incremental evolving domain adaptation , author=. IEEE Transactions on Knowledge and Data Engineering , volume=. 2016 , publisher=

    2016

  27. [35]

    arXiv preprint arXiv:2007.01807 , year=

    Continuously indexed domain adaptation , author=. arXiv preprint arXiv:2007.01807 , year=

    2007 arXiv

  28. [36]

    Advances in Neural Information Processing Systems , volume =

    Training for the Future: A Simple Gradient Interpolation Loss to Generalize Along Time , author =. Advances in Neural Information Processing Systems , volume =

  29. [37]

    International Conference on Learning Representations , year =

    Temporal Domain Generalization with Drift-Aware Dynamic Neural Networks , author =. International Conference on Learning Representations , year =

  30. [38]

    Advances in Neural Information Processing Systems , volume =

    Continuous Temporal Domain Generalization , author =. Advances in Neural Information Processing Systems , volume =

  31. [39]

    Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , year =

    Scaling Up Temporal Domain Generalization via Temporal Experts Averaging , author =. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , year =

    2025

  32. [40]

    Neural Computation , volume =

    Adjusting the Outputs of a Classifier to New a Priori Probabilities: A Simple Procedure , author =. Neural Computation , volume =

  33. [41]

    Advances in Neural Information Processing Systems , volume =

    Wild-Time: A Benchmark of In-the-Wild Distribution Shift over Time , author =. Advances in Neural Information Processing Systems , volume =

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.