Pith. sign in

REVIEW 4 major objections 5 minor 4 cited by

Financial Fine-tuning a Large Time Series Model

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Fine-tuning the TimesFM time-series foundation model on financial price data turns it from a predictor that often misses chance into one that beats chance and an AR(1) baseline.

desk verdict Fine-tuning TimesFM on financial data is a clean, reproducible empirical recipe, and the accuracy gains over the vanilla model look real, but the headline Sharpe 1.68 is not established: it rests on one overlapping year of zero-cost mock trading with no significance tests. read the letter →

arxiv 2412.09880 v1 pith:2THMFVP6 submitted 2024-12-13 q-fin.CP cs.LG

Xinghong Fu , Masanori Hirano , Kentaro Imajo This is my paper · ORCID
classification q-fin.CPcs.LG
keywords financialfine-tuningtimeseriesfoundationmodelTimesFMpricepredictioncontinualpre-trainingmarket-neutraltradingSharperatiolog-transformedMSEloss
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

Fine-tuning the pre-trained time-series foundation model TimesFM on financial price data changes it from a model that often underperforms random chance at price prediction into one that consistently beats chance and the original model across prediction horizons from 2 to 128 steps. The authors argue that the failure of the original model stems from the irregular, non-stationary character of price series, which differs sharply from the regular, seasonal time series used in pre-training. Their fine-tuning recipe — continual pre-training on roughly 100 million price points with a log-transformed mean-squared-error loss and random context masking — is what carries the improvement. They support this with mock trading: a market-neutral strategy based on 128-step predictions earns a 1.68 annualized Sharpe ratio and 3.6% annual returns on S&P500 stocks in a zero-cost setting, the strongest result across the markets tested.

What carries the argument

The load-bearing mechanism is the decoder-only TimesFM transformer, which reads price history in patches of 32 time points and predicts the next 128 points, with random masking during training so that it learns to forecast from any context length between 128 and 512. On top of this, the paper applies a log transform to prices before computing the mean-squared-error loss, turning the loss into something close to percentage error and preventing the rare crash events that otherwise produce NaN loss and destabilize training. The model is continually pre-trained from TimesFM's public checkpoint using SGD with warmup and cosine decay on a dataset dominated by hourly cryptocurrencies and stocks, and the whole run completes in under an hour on eight V100 GPUs.

What would settle it

Evaluate the fine-tuned model on 2024–2025 data with realistic transaction costs and a 128-step market-neutral strategy: if the net annualized Sharpe ratio is not positive while directional accuracy stays above chance, the central claim of a durable fine-tuning edge fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that a general-purpose time-series foundation model can be repurposed into a financial price predictor by continued pre-training on price data, and that the resulting model outperforms both the unmodified TimesFM and a chance-level classifier on directional accuracy and macro F1 at every tested horizon (2, 4, 8, 16, 32, 64, 128). In a zero-cost mock trading exercise on S&P500 stocks, the market-neutral version of the strategy reaches an annualized Sharpe ratio of 1.68, annual returns of 3.6%, and a maximum drawdown of about -0.1% at the 128-step horizon. The same model is the only one among the fine-tuned model, original TimesFM, a random model, and an AR(1) model to achieve positive Sharpe ratios in all four markets tested (S&P500, TOPIX500, currencies, daily crypto), although it trails AR(1) on currencies and crypto.

Load-bearing premise

The evaluation assumes that one out-of-sample year (2023), the chosen asset universe and horizons, and zero-cost mock trading fairly represent how the model would actually perform when traded or applied to other periods, and that continual pre-training dominated by hourly crypto and stock prices transfers to daily instruments.

Editorial extensions

If this is right

  • The vanilla TimesFM cannot be used as-is for financial price prediction: it underperforms a chance-rate model at four of seven prediction horizons.
  • Continual pre-training on price data with a log-transformed MSE loss and random masking yields consistent accuracy and macro-F1 gains over the original model at every horizon from 2 to 128.
  • A market-neutral trading strategy using the 128-step predictions achieves a 1.68 annualized Sharpe ratio and 3.6% annual returns on S&P500 stocks in zero-cost mock trading.
  • The fine-tuned model is the only method among those compared with positive Sharpe ratios in every market tested, although it trails AR(1) on currencies and crypto.
  • The full fine-tuning run takes under an hour on eight V100 GPUs, showing that adapting a 200-million-parameter foundation model to a new domain can be cheap when the dataset is modest.

Reading between the lines

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

  • Beyond the paper: because the loss is computed on log prices, the model effectively optimizes percentage error; this may make the fine-tuned checkpoint transfer more readily to assets whose price scales differ by orders of magnitude without reweighting, a property the paper does not test directly.
  • Beyond the paper: the random-masking scheme means the model is trained to forecast from arbitrary context lengths between 128 and 512; an ablation that fixes context length could reveal whether this flexibility is what drives the gains over the original model, or whether the data alone does.
  • Beyond the paper: since the fine-tuning set is dominated by hourly cryptocurrency and stock data, the same recipe on a granularity-balanced or synthetic-augmented dataset — directions the paper lists as future work — could plausibly close the gap with AR(1) on currencies and crypto, which is a testable prediction.
  • Beyond the paper: the reported 2023-only evaluation leaves open whether the 1.68 Sharpe is a momentum-style edge or a year-specific beta; applying the market-neutral strategy to 2024 and 2025 data with costs would discriminate between these.
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

4 major / 5 minor

Summary. The paper proposes continual pre-training of the TimesFM time-series foundation model on a large corpus of financial price data, using a log-MSE loss and a random masking scheme to stabilize training. The fine-tuned model is evaluated on price-direction accuracy, Macro F1, and two mock trading strategies (basic and market-neutral) across S&P500 stocks, TOPIX500 stocks, currencies, and cryptocurrencies. The central claims are that fine-tuning improves prediction accuracy over both the original TimesFM and a chance-rate baseline, and that a market-neutral strategy on S&P500 stocks achieves an annualized Sharpe ratio of 1.68 and 3.6% annual returns in a zero-cost mock trading setting. The authors release code and model weights for reproducibility.

Significance. If the central claims hold, the paper would provide a useful demonstration that a general-purpose time-series foundation model can be adapted to financial price prediction through continual pre-training, with potential practical application in trading strategy construction. The paper has concrete strengths: it uses a temporal holdout (2023 onward) for testing to avoid look-ahead bias, proposes a principled log-transformation of the loss, publishes code and model weights, and reports multiple performance metrics across several markets. However, the statistical evidence is currently incomplete: the accuracy/F1 improvements are asserted without significance tests, and the headline trading result relies on a single out-of-sample year, overlapping positions, and zero-cost assumptions. The paper's own discussion also concedes that the proposed model does not consistently beat a simple AR1 model, which weakens the broad claims made in the abstract.

major comments (4)
  1. [Section V-B and Figures 3-4] The text states that the fine-tuned model "outperforms random chance on all prediction horizon, hinting at statistically significant performance" and that fine-tuning "significantly improves" accuracy, but no significance tests, confidence intervals, or error bars are reported anywhere in the paper. The accuracy and Macro F1 comparisons are single point estimates computed on one test set; because this is the central claim of the paper, the authors should provide formal paired tests (e.g., McNemar's test for the accuracy classifications, or a bootstrap over series/days with cluster-robust standard errors) and report effect sizes with uncertainties. Without this, the word "significant" is not supported.
  2. [Section V-D, Table III, and Table IV] The headline result (annualized Sharpe 1.68 and 3.6% annual returns at h=128) is computed from a single out-of-sample year (2023) using a daily PnL series with strongly overlapping positions: each day's return contains 127 open layers of prior trades, so the effective number of independent holding periods is roughly 252/128 ≈ 2. No Newey-West adjustment, block-bootstrap, or other overlap-aware inference is reported, and the paper does not test whether the Sharpe is statistically distinguishable from zero. The evaluation also assumes zero trading costs; Table III's "neutral cost" of 0.60% at h=128 implies that a round-trip cost of 60 basis points would eliminate the entire 3.6% return. Please add overlap-adjusted confidence intervals for the Sharpe and a cost-sensitivity analysis over realistic round-trip costs (e.g., 5, 10, 20, 50 bp), and present the result as a single-year observation rather than an expected performance estimate.
  3. [Abstract, Section III-A, Section IV-A, and Table I] The reported size of the continual pre-training dataset is inconsistent: the abstract says 100 million time points, Section III-A twice refers to "80M time points", Section IV-A says "more than 100K time series and 90M time points", and the sum of Table I is approximately 82 million time points. The authors must reconcile these numbers and state a single, correct figure in all locations, since the dataset description is a central part of the method.
  4. [Section VI and Abstract/Table IV] The paper's own Discussion admits that "we are unable to ascertain consistently better performance over just a simple AR1 model," and Table IV confirms that AR1 outperforms the proposed model on Currencies (Sharpe 0.88 vs 0.25) and Crypto Daily (0.17 vs 0.26). The abstract's claim that the fine-tuned model "outperforms various benchmarks in terms of returns, sharpe ratio, max drawdown and trading cost" is therefore not supported as stated. The authors should qualify the claim to the specific markets and horizons where the improvement holds, and provide a statistical comparison against AR1 (e.g., a paired test on the PnL series) to justify any comparative statement.
minor comments (5)
  1. [Abstract and throughout] The model name is misspelled as "TimeFM" in the abstract; the correct name is "TimesFM."
  2. [Section IV-C3] The notation "1/(h−1)T" is ambiguous; it should be written as 1/((h−1)T) to indicate the order size per asset.
  3. [Section V-D and Table III] The text says "using a horizon length of H" where it should be "h"; also, the "Neutral Cost" entries are negative for some horizons (e.g., -0.006 for h=4), which should be explained since a negative neutral cost means the strategy loses money even at zero cost.
  4. [Sections I, IV-C3, and VI] There are several typos and grammatical errors: "siginificantly" in the Figure 1 caption, "most oftenly" in Section I, "dailu" in Section IV-C3, and "seris" and "repitition" in Section VI. A thorough proofread is needed.
  5. [Section V-D] The repository URL is typeset as "timesfm fin" with a space; the correct project name should be given so that the reproducibility link is usable.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper reports empirical comparisons on held-out 2023 data, with no load-bearing claim that reduces to its inputs.

full rationale

The paper's central claim is an empirical evaluation: fine-tuning TimesFM on financial price data improves prediction accuracy, as measured on a test set consisting of data from 2023 onward. The fine-tuning loss is a log-transformed MSE (Equation 3), the masking is inherited from the external TimesFM work, and the trading strategy consumes the model's own predictions. None of these steps defines the reported improvement in terms of itself. The chance-rate baseline is computed from the test set's up/down ratio, but it is a benchmark rather than a fitted parameter used to produce the model's predictions; it does not force the model's accuracy. The trading strategy's market-neutral adjustment and 'neutral cost' are standard post-hoc metrics, not circular inputs. The paper also explicitly acknowledges in Section VI that it cannot ascertain consistently better performance over a simple AR1 model, which is an honest limitation statement rather than a circular justification. There is no load-bearing self-citation: the TimesFM reference [14] is an external model, not prior work by the present authors, and no uniqueness theorem or ansatz is smuggled in through a citation. Consequently, no specific reduction of a prediction to its own inputs can be exhibited, and the circularity score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a small set of data-hygiene assumptions: that the test period and universe are fair, that the public data are clean, and that the log-MSE training objective is appropriate. No invented entities and no parameter fitted to the test set appear. The model weights themselves are learned, which is the intended mechanism.

free parameters (3)
  • Context length c = 512
    Maximum input length used at inference in mock trading; chosen for consistency with training, influences predictions.
  • Min context length = 128
    Minimum context length during random masking in training; chosen to ensure meaningful inputs.
  • Horizon length h = 2 to 128
    Trading and accuracy are evaluated across this sweep; performance peaks at h=128 in the S&P500 test.
assumptions (4)
  • domain assumption Historical data before 2023 are representative of 2023 market conditions for evaluation purposes.
    The model is fine-tuned on data up to 2022 and tested on 2023 only; no multiple market regimes are considered.
  • domain assumption Current S&P500 and TOPIX500 constituent lists do not introduce survivorship bias in the mock trading test.
    The paper does not describe adjusting constituents for the test period; using current members for past backtesting can inflate returns.
  • domain assumption The public API data from Yahoo Finance and Binance are accurate and correctly merged.
    No data validation or adjustment for splits and dividends is described for the price series.
  • ad hoc to paper MSE loss on log-transformed prices is an appropriate training objective for predicting price direction.
    The loss modification is a modeling choice introduced for this paper; its equivalence to percentage error only holds for small moves.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Financial Fine-tuning a Large Time Series Model." pith.science (2026). https://pith.science/paper/2THMFVP6

@misc{pith2026241209880,
  author       = {Pith},
  title        = {Pith review of: Financial Fine-tuning a Large Time Series Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2THMFVP6}},
  note         = {Machine review of arXiv:2412.09880}
}
read the original abstract

Large models have shown unprecedented capabilities in natural language processing, image generation, and most recently, time series forecasting. This leads us to ask the question: treating market prices as a time series, can large models be used to predict the market? In this paper, we answer this by evaluating the performance of the latest time series foundation model TimesFM on price prediction. We find that due to the irregular nature of price data, directly applying TimesFM gives unsatisfactory results and propose to fine-tune TimeFM on financial data for the task of price prediction. This is done by continual pre-training of the latest time series foundation model TimesFM on price data containing 100 million time points, spanning a range of financial instruments spanning hourly and daily granularities. The fine-tuned model demonstrates higher price prediction accuracy than the baseline model. We conduct mock trading for our model in various financial markets and show that it outperforms various benchmarks in terms of returns, sharpe ratio, max drawdown and trading cost.

Figures

Figures reproduced from arXiv: 2412.09880 by the authors.

Figure 1
Figure 1. We show that the baseline foundation time-series model TimesFM [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Training and Validation loss curves for fine tuning TimesFM following [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Accuracy score of fine-tuned versus original TimesFM and a chance [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Macro F1 score of fine-tuned versus original TimesFM, when [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Realized PnL using fine-tuned TimesFM traded on S&P500 stocks using the basic strategy, assuming no trading costs. Using the basic strategy, we see consistently positive gains over each horizon length at the end of the trading period. Note that using a horizon length o…
Figure 7
Figure 7. Figure 7: Realized PnL comparison between various models traded on S&P500 [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. CALM: A Framework for Continuous, Adaptive, and LLM-Mediated Anomaly Detection in Time-Series Streams

    cs.LG 2025-08 reject novelty 6.0 of 10

    CALM uses an LLM-as-a-Judge to curate anomalies for continuous fine-tuning of a time-series foundation model, improving anomaly detection on held-out stream segments.

  2. Time Series Foundation Models for Multivariate Financial Time Series Forecasting

    q-fin.GN 2025-07 reject novelty 6.0 of 10

    Pretrained TTM shows large transfer and sample-efficiency gains in three financial forecasting tasks relative to training from scratch, but methodological flaws including possible look-ahead bias weaken the quantitati...

  3. When Directional Accuracy Lies: A Base-Rate-Honest Benchmark for LoRA-Adapted TimesFM on Equity Forecasting

    q-fin.ST 2026-07 conditional novelty 5.0 of 10

    LoRA-fine-tuned TimesFM has no directional skill over the always-up base rate on NASDAQ-100 and S&P 500; its only benefit is slightly lower point-forecast error.

  4. Benchmarking Pre-Trained Time Series Models for Electricity Price Forecasting

    cs.LG 2025-06 conditional novelty 5.0 of 10

    No time series foundation model statistically outperforms the biseasonal MSTL model in most European day-ahead electricity price markets in 2024, though Chronos-Bolt and Time-MoE match traditional methods.

Reference graph

Works this paper leans on

42 extracted references · 19 canonical work pages · cited by 4 Pith papers

  1. [1]

    General exponential smoothing and the equivalent arma process,

    E. McKenzie, “General exponential smoothing and the equivalent arma process,” Journal of Forecasting, vol. 3, no. 3, pp. 333–344, 1984

    work page 1984

  2. [2]

    G. E. Box, G. M. Jenkins, G. C. Reinsel, and G. M. Ljung, Time series analysis: forecasting and control . John Wiley & Sons, 1970

    work page 1970

  3. [3]

    N-beats: Neural basis expansion analysis for interpretable time series forecasting,

    B. N. Oreshkin, D. Carpov, N. Chapados, and Y . Bengio, “N-beats: Neural basis expansion analysis for interpretable time series forecasting,” International Conference on Learning Representations (ICLR) , 2020

    work page 2020

  4. [4]

    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,” International Conference on Learning Representations (ICLR) , 2022

    work page 2022

  5. [5]

    Attention Is All You Need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention Is All You Need,” inAdvances in Neural Information Processing Systems , vol. 30, 2017, pp. 5999–6009

    work page 2017

  6. [6]

    Bert: Pre-training of deep bidirectional transformers for language understanding,

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre-training of deep bidirectional transformers for language understanding,” Association for Computational Linguistics , 2018

    work page 2018

  7. [8]

    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,” Neural Information Processing Systems, 2023

    work page 2023

  8. [9]

    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,” International Conference of Learning Representations , 2024. [Online]. Available: https://arxiv.org/abs/2310.01728

    arXiv 2024

Show all 42 references
  1. [10]

    Large language models for time series: A survey,

    X. Zhang, R. R. Chowdhury, R. K. Gupta, and J. Shang, “Large language models for time series: A survey,” International Joint Conference on Artificial Intelligence , 2024. [Online]. Available: https://arxiv.org/abs/2402.01801

    2024 arXiv

  2. [11]

    Llm processes: Numerical predictive distributions conditioned on natural language,

    J. Requeima, J. Bronskill, D. Choi, R. E. Turner, and D. Duvenaud, “Llm processes: Numerical predictive distributions conditioned on natural language,” 2024. [Online]. Available: https://arxiv.org/abs/2405.12856

    2024 arXiv

  3. [12]

    Test: Text prototype aligned embedding to activate llm’s ability for time series,

    C. Sun, H. Li, Y . Li, and S. Hong, “Test: Text prototype aligned embedding to activate llm’s ability for time series,” International Conference of Learning Representations , 2024. [Online]. Available: https://arxiv.org/abs/2308.08241

    2024 arXiv

  4. [13]

    Are language models actually useful for time series forecasting?

    M. Tan, M. A. Merrill, V . Gupta, T. Althoff, and T. Hartvigsen, “Are language models actually useful for time series forecasting?” 2024. [Online]. Available: https://arxiv.org/abs/2406.16964

    2024 arXiv

  5. [14]

    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,” International Conference of Machine Learning, 2024. [Online]. Available: https://arxiv.org/abs/2310.10688

    2024 arXiv

  6. [15]

    Generalized autoregressive conditional heteroskedastic- ity,

    T. Bollerslev, “Generalized autoregressive conditional heteroskedastic- ity,” Journal of Econometrics , vol. 31, no. 3, pp. 307–327, 1986. [Online]. Available: https://www.sciencedirect.com/science/article/pii/ 0304407686900631

    1986

  7. [16]

    A new approach to linear filtering and prediction problems,

    R. E. Kalman, “A new approach to linear filtering and prediction problems,” Transactions of the ASME–Journal of Basic Engineering , 1960

    1960

  8. [17]

    Hidden markov models for stock market prediction,

    L. Catello, L. Ruggiero, L. Schiavone, and M. Valentino, “Hidden markov models for stock market prediction,” 2023. [Online]. Available: https://arxiv.org/abs/2310.03775

    2023

  9. [18]

    Learning internal representations by error propagation,

    D. E. Rumelhart, G. E. Hinton, and R. J. Williams, “Learning internal representations by error propagation,” Readings in cognitive science ,

  10. [19]

    Long short-term memory,

    S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Comput., vol. 9, no. 8, p. 1735–1780, nov 1997. [Online]. Available: https://doi.org/10.1162/neco.1997.9.8.1735

    1997 doi

  11. [20]

    Stock market trend analysis using hidden markov model and long short term memory,

    M. Liu, J. Huo, Y . Wu, and J. Wu, “Stock market trend analysis using hidden markov model and long short term memory,” 2021. [Online]. Available: https://arxiv.org/abs/2104.09700

    2021 arXiv

  12. [21]

    Language models are few-shot learners,

    T. B. Brown, B. Mann, N. Ryder, M. Subbiah, J. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell, S. Agarwal, A. Herbert-V oss, G. Krueger, T. Henighan, R. Child, A. Ramesh, D. M. Ziegler, J. Wu, C. Winter, C. Hesse, M. Chen, E. Sigler, M. Litwin, S. Gray, B....

  13. [22]

    Llama: Open and efficient foundation language models,

    H. Touvron, T. Lavril, G. Izacard, X. Martinet, M.-A. Lachaux, T. Lacroix, B. Rozi `ere, N. Goyal, E. Hambro, F. Azhar, A. Rodriguez, A. Joulin, E. Grave, and G. Lample, “Llama: Open and efficient foundation language models,” 2023. [Online]. Available: https: //arxiv.org/abs/2...

    2023 arXiv

  14. [23]

    Mixtral of experts,

    A. Q. Jiang, A. Sablayrolles, A. Roux, A. Mensch, B. Savary, C. Bamford, D. S. Chaplot, D. de las Casas, E. B. Hanna, F. Bressand, G. Lengyel, G. Bour, G. Lample, L. R. Lavaud, L. Saulnier, M.-A. Lachaux, P. Stock, S. Subramanian, S. Yang, S. Antoniak, T. L. Scao, T. Gervet, T...

    2024 arXiv

  15. [24]

    Available: https://arxiv.org/abs/2005.14165

    [Online]. Available: https://arxiv.org/abs/2005.14165

    2005 arXiv

  16. [25]

    Vivit: A video vision transformer,

    A. Arnab, M. Dehghani, G. Heigold, C. Sun, M. Lu ˇci´c, and C. Schmid, “Vivit: A video vision transformer,” 2021. [Online]. Available: https://arxiv.org/abs/2103.15691

    2021 arXiv

  17. [26]

    Scalable diffusion models with transformers,

    W. Peebles and S. Xie, “Scalable diffusion models with transformers,”

  18. [27]

    An image is worth 16x16 words: Trans- formers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Trans- formers for image recognition at scale,” Proceedings of the International Conf...

    2021

  19. [28]

    Fingpt: Open-source financial large language models,

    H. Yang, X.-Y . Liu, and C. D. Wang, “Fingpt: Open-source financial large language models,” 2023. [Online]. Available: https: //arxiv.org/abs/2306.06031

    2023

  20. [29]

    Darts: User-friendly modern machine learning for time series,

    J. Herzen, F. L ¨assig, S. G. Piazzetta, T. Neuer, L. Tafti, G. Raille, T. V . Pottelbergh, M. Pasieka, A. Skrodzki, N. Huguenin, M. Dumonal, J. Ko ´scisz, D. Bader, F. Gusset, M. Benheddi, C. Williamson, M. Kosinski, M. Petrik, and G. Grosch, “Darts: User-friendly modern mach...

    2022 arXiv

  21. [30]

    Monash time series forecasting archive,

    R. Godahewa, C. Bergmeir, G. I. Webb, R. J. Hyndman, and P. Montero-Manso, “Monash time series forecasting archive,” 2021. [Online]. Available: https://arxiv.org/abs/2105.06643

    2021 arXiv

  22. [31]

    Timegpt-1,

    A. Garza, C. Challu, and M. Mergenthaler-Canseco, “Timegpt-1,” 2024. [Online]. Available: https://arxiv.org/abs/2310.03589

    2024 arXiv

  23. [32]

    On model evaluation under non-constant class imbalance,

    J. Brabec, T. Kom ´arek, V . Franc, and L. Machlica, “On model evaluation under non-constant class imbalance,” 2020. [Online]. Available: https://arxiv.org/abs/2001.05571

    2020 arXiv

  24. [33]

    Macro f1 and macro f1,

    J. Opitz and S. Burst, “Macro f1 and macro f1,” 2021. [Online]. Available: https://arxiv.org/abs/1911.03347

    2021 arXiv

  25. [34]

    Synthetic data from diffusion models improves imagenet classification,

    S. Azizi, S. Kornblith, C. Saharia, M. Norouzi, and D. J. Fleet, “Synthetic data from diffusion models improves imagenet classification,”

  26. [35]

    Informer: Beyond efficient transformer for long sequence time-series forecasting,

    H. Zhou, S. Zhang, J. Peng, S. Zhang, J. Li, H. Xiong, and W. Zhang, “Informer: Beyond efficient transformer for long sequence time-series forecasting,” Proceedings of the AAAI conference on artificial intelli- gence, 2021

    2021

  27. [36]

    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,”

  28. [37]

    Fine-tuning can distort pretrained features and underperform out-of-distribution,

    A. Kumar, A. Raghunathan, R. M. Jones, T. Ma, and P. Liang, “Fine-tuning can distort pretrained features and underperform out-of-distribution,” in International Conference on Learning Representations , 2022. [Online]. Available: https: //openreview.net/forum?id=UYneFzXSJWh

    2022

  29. [38]

    Understanding intermediate layers using linear classifier probes,

    G. Alain and Y . Bengio, “Understanding intermediate layers using linear classifier probes,” 2018. [Online]. Available: https://arxiv.org/abs/ 1610.01644

    2018 arXiv

  30. [39]

    Available: https://arxiv.org/abs/2304.08466

    [Online]. Available: https://arxiv.org/abs/2304.08466

    arXiv

  31. [40]

    How transferable are video representations based on synthetic data?

    Y .-w. Kim, S. Mishra, S. Jin, R. Panda, H. Kuehne, L. Karlinsky, V . Saligrama, K. Saenko, A. Oliva, and R. Feris, “How transferable are video representations based on synthetic data?” in Advances in Neural Information Processing Systems , S. Koyejo, S. Mohamed, A. Agarwal, D...

    2022

  32. [1986]

    Available: https://api.semanticscholar.org/CorpusID: 62245742

    [Online]. Available: https://api.semanticscholar.org/CorpusID: 62245742

  33. [2021]

    Available: https://arxiv.org/abs/2106.09685

    [Online]. Available: https://arxiv.org/abs/2106.09685

    arXiv

  34. [2023]

    Available: https://arxiv.org/abs/2212.09748

    [Online]. Available: https://arxiv.org/abs/2212.09748

    arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.