FedChronos: Federated Fine-Tuning of Time-Series Foundation Models for Privacy-Preserving Commodity Price Forecasting

Akramul Azim; Amit Sharma; Nitin Auluck

REVIEW 4 major objections 5 minor 33 references

FedChronos: Federated Fine-Tuning of Time-Series Foundation Models for Privacy-Preserving Commodity Price Forecasting

T0 review · 4 major / 5 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 Differential privacy noise, added per round during federated LoRA fine-tuning of a pre-trained time-series model, cuts mean absolute percentage error by 31% over zero-shot while bounding information leakage.

desk verdict A useful gap and an honest but under-tested claim: FedChronos shows DP noise can regularize federated TSFM fine-tuning, but the causal story needs a non-private control before believing the strong conclusion. read the letter →

arxiv 2608.01290 v1 pith:46LI3YXS submitted 2026-08-02 cs.LG cs.DC

Amit Sharma , Nitin Auluck , Akramul Azim This is my paper · ORCID

classification cs.LGcs.DC

keywords federatedlearningparameter-efficientfine-tuningLoRAtime-seriesfoundationmodelsdifferentialprivacycommoditypriceforecastingnon-IIDdataFedProx

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

The paper tries to establish that federated fine-tuning of an already pre-trained time-series foundation model is viable on small, institutionally fragmented datasets only when the adaptation is regularized, and that per-round differential privacy noise can supply that regularization. On daily potato and onion price series from 15 Indian wholesale markets, naive LoRA fine-tuning of the frozen Chronos-T5-Tiny backbone overfits and falls below zero-shot accuracy. Adding client-level differential privacy at a per-round budget of $\varepsilon=5$ cuts mean absolute percentage error to 69.78%, a 31% improvement over zero-shot and a 26% improvement over the best pooled baseline, while the per-round model exchange shrinks 86-fold because only 384 KB of adapter weights travel. The broader claim is that privacy and accuracy can be complementary in this setting rather than competing.

What carries the argument

The load-bearing mechanism is the combination of a frozen pre-trained T5 encoder-decoder backbone with LoRA adapters on its query and value projections, trained across clients and aggregated by FedAvg or FedProx, with client-level differential privacy implemented by clipping each round's adapter update to an $\ell_2$ ball of radius $C=1$ and adding Gaussian noise calibrated to a per-round $(\varepsilon,\delta)$ target. With only 98,304 trainable parameters, the per-parameter signal-to-noise ratio at a given $\varepsilon$ is higher than for full-model updates, which the paper argues amplifies the regularizing effect of the DP noise. The mechanism accounts both for the overfitting failure, unregularized adapters fitting noise in a 98,304-dimensional subspace, and for the recovery at $\varepsilon=5$, where noise keeps the adapters from settling into sharp minima.

What would settle it

Run the same 15-client federated LoRA setup with several random seeds at $\varepsilon=5$, and also run a non-private early-stopping baseline and a control with unclipped noise of the same magnitude. If early stopping alone reaches the same 69.78% MAPE, or if the $\varepsilon=5$ gain does not replicate under a fixed test set across seeds, the paper's central claim that DP noise is the regularizing mechanism would be falsified.

Watch

Extended reading notes

Core claim

FedChronos discovers that differential privacy noise does not simply cost accuracy in federated time-series foundation model fine-tuning; at the right calibration it prevents the overfitting caused by tuning 98,304 LoRA parameters on roughly 515 training observations per commodity series. Without DP, every fine-tuning variant tested (centralized, FedAvg, FedProx, local-only) ends up worse than the zero-shot MAPE of 101.76%, with centralized LoRA worst at 140.33%. With per-round client-level DP at $\varepsilon=5$, MAPE drops to 69.78%, below the best pooled baseline LSTM at 94.97%, and the privacy-accuracy curve is U-shaped: $\varepsilon=8$ gives too little regularization (90.41%) and $\varepsilon=3$ gives too much noise (88.64%). The paper flags that the exact optimum is dataset-dependent and that the reported $\varepsilon$ is per-round, not composed across the 50 training rounds.

Load-bearing premise

The claim that differential privacy is the cause of the accuracy gain rests on a single run for each configuration, a per-round privacy setting of $\varepsilon=5$ tuned on the same test set that produced the headline numbers, and no non-private control such as early stopping or equal-magnitude unclipped noise.

Editorial extensions

If this is right

Federated time-series foundation model fine-tuning on small client datasets should not be run without a regularizer: unregularized LoRA produces models worse than the pre-trained zero-shot model.
A suitable per-round privacy budget can simultaneously bound information leakage and improve forecast accuracy, so privacy and accuracy need not be traded off in this regime.
The 86× communication reduction from exchanging LoRA adapters (384 KB per round versus 33 MB for the full model) makes federated TSFM adaptation practical on constrained rural or edge links.
FedProx narrows the per-client accuracy spread relative to FedAvg, but the DP regularization effect is an order of magnitude larger than the difference between aggregation strategies.
The reported privacy guarantee is per-round; composing it across all 50 rounds would yield a larger cumulative $\varepsilon$, so end-to-end privacy accounting remains an open requirement.

Reading between the lines

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

Beyond the paper: if the regularizing effect is generic, the same recipe of a frozen time-series foundation model plus LoRA plus per-round DP noise could turn privacy budgets into a tunable regularizer for other siloed time-series domains, such as energy load or clinical monitoring.
Beyond the paper: because no non-private control was run, the specific attribution of the gain to differential privacy rather than to noise in general or to implicit early stopping is not settled; a seed sweep with an early-stopping baseline and an unclipped matched-noise control would discriminate.
Beyond the paper: the U-shaped accuracy curve suggests the privacy parameter could be selected on validation data in practice, effectively treating the privacy budget as a hyperparameter of the forecasting system.
Beyond the paper: pairing secure aggregation with per-round DP would hide individual adapters from the server, but the reported per-round $\varepsilon$ would still need end-to-end composition before the guarantee is meaningful in deployment.

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.

Desk Editor's Note

read the letter

FedChronos tackles a real gap — federated fine-tuning of an already pre-trained time-series foundation model via LoRA — and the empirical story is more interesting than the usual "we applied known components" paper. The overfitting observation (naive LoRA fine-tuning on small per-client data drops below zero-shot) is a genuinely useful data point, and the DP-as-regularizer result at eps=5 is suggestive. The paper is honest: it reports all three metrics even when they disagree, flags that RMSE is best at eps=8, and explicitly lists early stopping as a missing control. That honesty earns credit.

What's new: prior federated TSFM work either pre-trains from scratch or aligns prototypes; nobody has adapted a fixed pre-trained TSFM with LoRA in a federated loop. The 86x communication reduction is real and makes edge deployment plausible. The framework is clearly described and the experimental setup uses a natural non-IID dataset.

The weak spot is the central causal claim. "DP noise acts as implicit regularization" is not actually tested against a non-private regularizer. Early stopping at the validation minimum (which the paper's own Figure 4 suggests would work) could produce the same MAPE, and the stress-test note correctly says the eps=5 optimum was selected on the same test set used for the headline numbers, from a single run. So the U-shaped curve is a selected maximum, not a measured optimum. That doesn't kill the paper, but it demotes the key finding from "privacy and accuracy are complementary" to "some form of regularization is needed for small-data LoRA fine-tuning; DP noise can serve that role." The abstract overreaches slightly.

Minor but worth noting: the DP guarantee is per-round, not composed over 50 rounds, so the privacy claim in the abstract needs a caveat; the dataset covers two commodities and 15 mandis in simulation; no code is released yet. These are acknowledged in the limitations, but they matter for reproducibility.

Who this is for: people working on federated TSFM fine-tuning, parameter-efficient adaptation on edge devices, and the privacy-regularization interplay. It deserves a serious referee — the gap is real, the experiments are honestly reported, and the limitations are stated. But the referee should ask for an early-stopping baseline, multiple seeds, and a clearer privacy accounting before accepting the strong conclusion.

Referee Report

4 major / 5 minor

Summary. FedChronos proposes federated LoRA fine-tuning of a pre-trained Chronos-T5-Tiny time-series foundation model on daily commodity price data from 15 Indian mandis, using FedAvg and FedProx aggregation with optional per-round client-level differential privacy. The paper reports that unregularized LoRA fine-tuning overfits on small per-client datasets and falls below the zero-shot baseline (MAPE around 101.76%), while adding DP at ε=5 yields MAPE 69.78%, a 31% improvement over zero-shot and 26% over the best traditional baseline. It also reports an 86× reduction in per-round communication cost through LoRA adapters. The authors explicitly flag several limitations: the results come from a single run per configuration, ε is tuned on the same test set used for the headline numbers, the privacy guarantee is per-round rather than end-to-end, and no non-private early-stopping control was tested.

Significance. If the central empirical claim were established, the paper would make a useful contribution to federated time-series foundation model adaptation: it identifies a concrete overfitting failure mode for small federated LoRA fine-tuning, provides a naturally non-IID agricultural benchmark, and demonstrates a large communication reduction. The paper is also commendably honest in reporting its limitations, including the RMSE inconsistency at ε=5 and the absence of a non-private regularization control. However, the headline claim that DP noise specifically acts as implicit regularization is not yet supported by the experimental design: there is no controlled comparison against early stopping or matched-magnitude non-private noise, each configuration was run once, and ε was selected on the same test data used to report the improvement. The significance of the result therefore depends on additional experiments that the authors themselves identify as necessary.

major comments (4)

[Section V-D, Table V, Section VI-B] The causal claim that DP noise, rather than generic regularization, drives the accuracy gain is not supported by a controlled comparison. Figure 4 shows that validation loss for FedAvg and FedProx reaches a minimum around rounds 5–7 and rises thereafter, yet no early-stopping baseline without DP is reported. A simple no-DP FedAvg or FedProx run stopped at the validation minimum would test directly whether the observed MAPE reduction is specific to differential privacy or is just an instance of regularization preventing overfitting. Until such a control is provided, the paper's conclusion that 'privacy and accuracy can be complementary' is stronger than the evidence supports.
[Section IV-D, Section VI-B, Table V] The ε=5 optimum is selected from the same test data used to compute the headline 31% improvement, and each configuration is run only once. This means the reported gain is a selected maximum with no confidence interval, and the absence of seed variation makes it impossible to assess whether ε=5 is a stable optimum or a single favorable draw. The internal inconsistency in Table V, where RMSE is best at ε=8 while MAE and MAPE favor ε=5, further weakens the impression of a robust optimum. The authors should report multiple seeds, select ε on validation data (or otherwise account for selection), and provide per-seed spreads for all metrics.
[Section V-D, Tables IV and V] The comparison between the no-DP row of Table V and the unregularized results in Table IV is not anchored. Table IV reports FedAvg LoRA MAPE of 139.60% and FedProx LoRA MAPE of 134.82%, while Table V lists 'No DP (ε=∞)' as MAPE 134.73%. It is unclear which aggregation strategy corresponds to the no-DP baseline in Table V, and this ambiguity matters because the claimed improvement from DP is measured against that baseline. The authors should state explicitly which federated protocol is used for the no-DP row and why its value differs from the corresponding entry in Table IV.
[Section III-E, Abstract] The privacy guarantee is per-round: the reported ε∈{3,5,8} is calibrated for a single round's update, and composition across R=50 rounds is not performed. The paper states this limitation clearly in Section III-E and Section VI-B, but the abstract and title present 'privacy-preserving' and ε=5 without this qualifier. To avoid overstating the guarantee, the abstract should either state that ε is a per-round budget or the paper should provide an end-to-end accounting via Rényi DP or a moments accountant.

minor comments (5)

[Section IV-D, Figure 6] Figure 6 draws a smooth U-shaped curve through only three ε values plus a no-DP point; without error bars or additional ε values, the visual suggestion of a well-defined optimum near ε=5 is not supported by the data. A scatter plot with per-seed points, or error bars if seeds are added, would be more appropriate.
[Section III-E] The claim that LoRA gives a higher per-parameter signal-to-noise ratio than full-model DP-FL because noise is distributed over fewer parameters is heuristic and not derived. Since the noise variance is per-coordinate for a fixed σ, the comparison depends on the relative magnitudes of the coordinate-wise updates, which are not reported. This statement should be softened or supported.
[Section IV-A, Table II] The dataset description says approximately 515 training observations per commodity series after filtering, but the selection of 15 mandis from 464 is not fully specified. It would help to report the distribution of series lengths and the exact criteria used to choose these particular clients, since the non-IID characterization is central to the experimental setting.
[Throughout] There are several LaTeX rendering artifacts, such as 'na ¨ıve' in the abstract and Section V, and the proximal loss notation in Equation (4). These should be cleaned up before publication.
[Algorithm 1, Section III-E] The paper operates under an honest-but-curious threat model but does not discuss secure aggregation; without it, the server observes individual client adapters. This is acceptable given the threat model, but it should be stated explicitly in the privacy discussion so readers do not infer a stronger guarantee than is provided.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper reports empirical measurements; test-set ε selection and missing non-private controls are validity concerns, not definitional circularity.

full rationale

The paper's load-bearing claims are empirical measurements rather than derivations, and none of the defined circularity patterns applies. The headline result (ε=5 gives MAPE 69.78%, a 31% improvement over zero-shot) is a directly reported experimental observation from Table V, not a quantity derived from a definition that already contains it. The DP noise is calibrated with the standard external Gaussian-mechanism formula σ = sqrt(2 ln(1.25/δ))/ε, not with a formula fitted to the outcome. No self-citations are used; all substantive prior results cited (Chronos, LoRA, FedAvg, FedProx, DP) are external, and none is invoked as a load-bearing uniqueness or equivalence argument. The ε sweep over {3,5,8} and the selection of ε=5 on the test set is an acknowledged model-selection choice, and the authors explicitly frame it as dataset-dependent guidance rather than a predicted optimum: 'we tune ε on this dataset, so the specific optimum near ε=5 is best read as dataset-dependent guidance rather than a universal setting' (Section VI-B). They also explicitly flag the single-run nature and the missing non-private early-stopping control: 'The numbers here come from a single run per configuration, so confirming the U-shaped ε curve across seeds, with confidence intervals, and adding a non-private control ... would pin down how much of the gain is specific to differential privacy rather than to regularization in general.' These admissions are limitations on causal attribution and statistical robustness, not evidence that any result is equivalent to its input by construction. The regularizer-vs-DP-noise question is a scientific counterfactual that the paper does not attempt to settle by definition, and the absence of such a control is a correctness risk, not circularity.

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

The central empirical claims rest on standard privacy math plus domain assumptions about the data, client selection, and single-run stability. The paper itself flags the per-round nature of epsilon and the absence of seeds and controls. No new entities are introduced; the framework is a composition of known components.

free parameters (5)

Privacy budget epsilon = 5 (selected as best among {3, 5, 8})
Chosen from test-set MAPE sweep; the headline 31% improvement is reported at the selected value, which inflates expected out-of-sample performance.
Clipping norm C = 1.0
Fixed by hand; the DP guarantee and noise magnitude scale with C, so it affects both privacy and the regularization strength.
LoRA rank r = 8
Chosen by hand; determines adapter size (98,304 parameters), communication cost, and capacity for overfitting. Rank 4/16/32 appear only in the communication table, not in accuracy experiments.
FedProx proximal coefficient mu = 0.01
Chosen for the FedProx comparison; the paper concludes the DP effect dominates the FedAvg-vs-FedProx gap, so the exact mu is not central, but it is a hand-set value.
Training schedule R and E = R=50, E=3
The protocol runs 50 communication rounds with 3 local epochs and no early stopping; this schedule contributes to the observed overfitting and is a chosen setting.

assumptions (5)

standard math The Gaussian mechanism with sigma = sqrt(2 ln(1.25/delta))/epsilon calibrates the per-round privacy guarantee for the LoRA update.
Used in Section III-E and IV-D to set sigma for epsilon in {3, 5, 8}; this is standard DP math, but it is applied per round without composition across the 50 rounds, so the end-to-end guarantee is weaker than the reported epsilon.
domain assumption Client-level DP protects the entire local dataset against a single-round adversary.
The threat model in Section III-E assumes honest-but-curious observers and treats the protected unit as the client's dataset; without composition or secure aggregation the guarantee for the full training run is not established.
domain assumption The 15 selected mandis and two commodities are representative of the federated TSFM fine-tuning setting.
Section IV-B selects clients spanning 9 states but gives no rule for choosing them; generalization to other markets, commodities, and data qualities is assumed.
domain assumption Preprocessing (daily resampling, interpolation of gaps up to 7 days, density filters) does not materially distort the price series.
Section IV-A describes these steps; if interpolation injects smooth structure, the measured overfitting and the DP-regularization effect could be artifacts of the preprocessing.
domain assumption Validation loss is a reliable proxy for generalization, and training for 50 rounds without early stopping is the appropriate protocol for diagnosing overfitting.
Fig. 4 uses the rise in validation loss to diagnose overfitting, but the final models are not selected at the validation minimum; comparing against early stopping is deferred to future work in Section VI-B.

how reviews work

0 comments

Cite this review

Pith. "Pith review of FedChronos: Federated Fine-Tuning of Time-Series Foundation Models for Privacy-Preserving Commodity Price Forecasting." pith.science (2026). https://pith.science/paper/46LI3YXS

@misc{pith2026260801290,
  author       = {Pith},
  title        = {Pith review of: FedChronos: Federated Fine-Tuning of Time-Series Foundation Models for Privacy-Preserving Commodity Price Forecasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/46LI3YXS}},
  note         = {Machine review of arXiv:2608.01290}
}

abstract

Time-series foundation models (TSFMs) such as Chronos have demonstrated strong forecasting capabilities across domains, yet adapting them to institutionally fragmented settings, where data cannot be centralized due to regulatory, competitive, or sovereignty constraints, remains unexplored. We introduce FedChronos, a framework for federated parameter-efficient fine-tuning of an already pre-trained TSFM, a setting that existing federated time-series work has not addressed, since prior methods either pre-train from scratch or align prototypes rather than adapt a fixed backbone. Our approach applies Low-Rank Adaptation (LoRA) to the Chronos-T5 backbone and trains across distributed clients using FedAvg and FedProx, transmitting only lightweight adapter weights (384~KB per round, an 86$\times$ reduction over full-model exchange). We evaluate FedChronos on daily commodity prices from 15 Indian agricultural markets across 9 states, a naturally non-IID federated setting, and find that na\"ive LoRA fine-tuning overfits substantially on small per-client datasets, dropping below zero-shot performance. We further observe that differential privacy (DP) noise can act as implicit regularization and counteract this overfitting: in our experiments the strongest configuration ($\varepsilon = 5$) reduces mean absolute percentage error (MAPE) by 31% over zero-shot and 26% over the best traditional baseline, while bounding each round's information leakage via per-round $(\varepsilon, \delta)$-differential privacy. Because the model is compact and the updates are small, the approach also suits edge AI deployments where both the network link and the client device are constrained. Overall, our findings suggest that privacy and accuracy can be complementary rather than competing objectives in federated TSFM fine-tuning.

Figures

Figures reproduced from arXiv: 2608.01290 by the authors.

**Figure 2.** Internal architecture of FedChronos at each client. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 5.** V. RESULTS We organize our experimental evaluation around four questions: (1) How do baselines and the pre-trained model perform on mandi data? (2) Does na¨ıve fine-tuning improve over zeroshot? (3) How do FL algorithms compare, and what role does non-IID heterogeneity play? (4) How does differential privacy TABLE III: Baseline forecasting performance on mandi price data. Method MAE RMSE MAPE (%) ARIMA 1597.87 163… view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Validation loss during federated training. Both Fe [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

**Figure 6.** Figure 6: MAPE as a function of privacy budget ε. The curve suggests an apparent optimum near ε = 5: too little noise (large ε) allows overfitting, while too much noise (small ε) overwhelms the learning signal. A plausible mechanistic explanation is the following. LoRA fine-tuni…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 18 canonical work pages

[1]

AI-based market intelligence systems for farmer collectives: A case study from India,

R. Ladhar, S. Sharma, S. Tangirala, N. Guptaet al., “AI-based market intelligence systems for farmer collectives: A case study from India,” ACM Journal on Computing and Sustainable Societies, 2023

work page 2023
[2]

The Digital Personal Data Protection Act, 2023,

Government of India, “The Digital Personal Data Protection Act, 2023,” Ministry of Electronics and Information Technology, 2023

work page 2023
[3]

Regulation (EU) 2016/679 — General Data Protection Regulation,

European Parliament and Council of the European Union, “Regulation (EU) 2016/679 — General Data Protection Regulation,” 2016

work page 2016
[4]

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. Kapooret al., “Chronos: Learning the language of time series,”arXiv preprint arXiv:2403.07815, 2024

arXiv 2024
[5]

Lag-Llama: Towards foundation models for probabilistic time series forecasting,

K. Rasul, A. Ashok, A. R. Williams, H. Ghonia, R. Bhagwatkar, A. Kho- rasaniet al., “Lag-Llama: Towards foundation models for probabilistic time series forecasting,”arXiv preprint arXiv:2310.08278, 2024

arXiv 2024
[6]

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,”arXiv preprint arXiv:2310.10688, 2024

arXiv 2024
[7]

TimeGPT-1,

A. Garza, C. Challu, and M. Mergenthaler-Canseco, “TimeGPT-1,” arXiv preprint arXiv:2310.03589, 2024

arXiv 2024
[8]

Attention is all you need,

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,” inAdvances in Neural Information Processing Systems (NeurIPS), 2017

2017

Show all 33 references

[9]

Communication-efficient learning of deep networks from decentralized data,

B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, “Communication-efficient learning of deep networks from decentralized data,” inInternational Conference on Artificial Intelligence and Statistics (AISTATS), 2017

2017
[10]

LoRA: Low-rank adaptation of large language models,

E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, and W. Chen, “LoRA: Low-rank adaptation of large language models,” inInternational Conference on Learning Representations (ICLR), 2022

2022
[11]

Federated adaptive fine-tuning of large language models with heterogeneous quantization and LoRA,

Z. Gao, Z. Zhang, Y . Guo, and Y . Gong, “Federated adaptive fine-tuning of large language models with heterogeneous quantization and LoRA,” inIEEE INFOCOM, 2025

2025
[12]

A survey on federated fine-tuning of large language models,

Y . Wu, C. Tian, J. Li, H. Sun, K.-H. Tam, and Z. Zhou, “A survey on federated fine-tuning of large language models,”arXiv preprint arXiv:2503.12016, 2025

2025
[13]

Exploring the limits of transfer learning with a unified text-to-text transformer,

C. Raffel, N. Shazeer, A. Roberts, K. Lee, S. Narang, M. Matena, Y . Zhou, W. Li, and P. J. Liu, “Exploring the limits of transfer learning with a unified text-to-text transformer,”Journal of Machine Learning Research, vol. 21, no. 140, pp. 1–67, 2020

2020
[14]

Federated optimization in heterogeneous networks,

T. Li, A. K. Sahu, M. Zaheer, M. Sanjabi, A. Talwalkar, and V . Smith, “Federated optimization in heterogeneous networks,” inConference on Machine Learning and Systems (MLSys), 2020

2020
[15]

Deep learning with differential privacy,

M. Abadi, A. Chu, I. Goodfellow, H. B. McMahan, I. Mironov, K. Tal- war, and L. Zhang, “Deep learning with differential privacy,” inACM Conference on Computer and Communications Security (CCS), 2016

2016
[16]

Federated foundation models on heterogeneous time series,

S. Chen, G. Long, J. Jiang, and C. Zhang, “Federated foundation models on heterogeneous time series,” inProceedings of the AAAI Conference on Artificial Intelligence, 2025

2025
[17]

Discrete prototypical memories for federated time series foundation models,

L. Deng, Q. Liu, X. Niu, S. Chen, S. Sun, Y . Wu, G. Long, and Y . Liang, “Discrete prototypical memories for federated time series foundation models,”arXiv preprint arXiv:2604.04475, 2026

2026 arXiv
[18]

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,” inInternational Conference on Learning Representations (ICLR), 2024

2024
[19]

Federated learning with non-IID data,

Y . Zhao, M. Li, L. Lai, N. Suda, D. Civin, and V . Chandra, “Federated learning with non-IID data,”arXiv preprint arXiv:1806.00582, 2018

2018 arXiv
[20]

QLoRA: Efficient finetuning of quantized language models,

T. Dettmers, A. Pagnoni, A. Holtzman, and L. Zettlemoyer, “QLoRA: Efficient finetuning of quantized language models,” inAdvances in Neural Information Processing Systems (NeurIPS), 2023

2023
[21]

Edge-FIT: Federated instruction tuning of quantized LLMs for privacy-preserving smart home environments,

V . Venkatesh, V . Kamanuru, L. Kumaret al., “Edge-FIT: Federated instruction tuning of quantized LLMs for privacy-preserving smart home environments,”arXiv preprint, 2025

2025
[22]

Agricultural data privacy and federated learning: A review of challenges and opportuni- ties,

R. Dembani, I. Karvelas, N. Akbar, S. Rizouet al., “Agricultural data privacy and federated learning: A review of challenges and opportuni- ties,”Computers and Electronics in Agriculture, 2025

2025
[23]

Federated learning-based approach for crop recommendation and market stability in agriculture,

S. Kumar, T. Maurya, M. Raiet al., “Federated learning-based approach for crop recommendation and market stability in agriculture,”Federated Learning for Agriculture, 2026

2026
[24]

VLLFL: A vision-language model based lightweight federated learning framework for smart agriculture,

L. Li, J. Li, D. Chen, L. Pu, H. Yao, and Y . Huang, “VLLFL: A vision-language model based lightweight federated learning framework for smart agriculture,”arXiv preprint arXiv:2504.13365, 2025

2025
[25]

AgriGen: A prompt-tuned, mul- tilingual LLM-based Q&A system for smarter agriculture,

V . Kamduri, P. Gupta, and C. El Kari, “AgriGen: A prompt-tuned, mul- tilingual LLM-based Q&A system for smarter agriculture,” inApplied Imagery Pattern Recognition Workshop (AIPR). Springer, 2026

2026
[26]

A review paper on the study of deep learning and machine learning models used in forecasting Indian crop prices,

S. Srivastava and S. Dahiya, “A review paper on the study of deep learning and machine learning models used in forecasting Indian crop prices,” inInternational Conference on Data Analytics. Springer, 2025

2025
[27]

Flower: A friendly federated learning research framework,

D. J. Beutel, T. Topal, A. Mathur, X. Qiu, J. Fernandez-Marques, Y . Gao, L. Sani, K. H. Li, T. Parcollet, P. P. B. de Gusm ˜ao, and N. D. Lane, “Flower: A friendly federated learning research framework,”arXiv preprint arXiv:2007.14390, 2020

2007 arXiv
[28]

The algorithmic foundations of differential privacy,

C. Dwork and A. Roth, “The algorithmic foundations of differential privacy,”Foundations and Trends in Theoretical Computer Science, vol. 9, no. 3–4, pp. 211–407, 2014

2014
[29]

R ´enyi differential privacy,

I. Mironov, “R ´enyi differential privacy,” inIEEE 30th Computer Security Foundations Symposium (CSF), 2017, pp. 263–275

2017
[30]

G. E. P. Box and G. M. Jenkins,Time Series Analysis: Forecasting and Control. San Francisco: Holden-Day, 1976

1976
[31]

Long short-term memory,

S. Hochreiter and J. Schmidhuber, “Long short-term memory,”Neural Computation, vol. 9, no. 8, pp. 1735–1780, 1997

1997
[32]

Adding gradient noise improves learning for very deep networks,

A. Neelakantan, L. Vilnis, Q. V . Le, I. Sutskever, L. Kaiser, K. Karol, and J. Martens, “Adding gradient noise improves learning for very deep networks,”arXiv preprint arXiv:1511.06807, 2015

2015 arXiv
[33]

Generalization in adaptive data analysis and holdout reuse,

C. Dwork, V . Feldman, M. Hardt, T. Pitassi, O. Reingold, and A. Roth, “Generalization in adaptive data analysis and holdout reuse,” inAd- vances in Neural Information Processing Systems (NeurIPS), 2015

2015

Pith tools

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

[1] [1]

AI-based market intelligence systems for farmer collectives: A case study from India,

R. Ladhar, S. Sharma, S. Tangirala, N. Guptaet al., “AI-based market intelligence systems for farmer collectives: A case study from India,” ACM Journal on Computing and Sustainable Societies, 2023

work page 2023

[2] [2]

The Digital Personal Data Protection Act, 2023,

Government of India, “The Digital Personal Data Protection Act, 2023,” Ministry of Electronics and Information Technology, 2023

work page 2023

[3] [3]

Regulation (EU) 2016/679 — General Data Protection Regulation,

European Parliament and Council of the European Union, “Regulation (EU) 2016/679 — General Data Protection Regulation,” 2016

work page 2016

[4] [4]

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. Kapooret al., “Chronos: Learning the language of time series,”arXiv preprint arXiv:2403.07815, 2024

arXiv 2024

[5] [5]

Lag-Llama: Towards foundation models for probabilistic time series forecasting,

K. Rasul, A. Ashok, A. R. Williams, H. Ghonia, R. Bhagwatkar, A. Kho- rasaniet al., “Lag-Llama: Towards foundation models for probabilistic time series forecasting,”arXiv preprint arXiv:2310.08278, 2024

arXiv 2024

[6] [6]

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,”arXiv preprint arXiv:2310.10688, 2024

arXiv 2024

[7] [7]

TimeGPT-1,

A. Garza, C. Challu, and M. Mergenthaler-Canseco, “TimeGPT-1,” arXiv preprint arXiv:2310.03589, 2024

arXiv 2024

[8] [8]

Attention is all you need,

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,” inAdvances in Neural Information Processing Systems (NeurIPS), 2017

2017