Transformers with Sparse Attention for Granger Causality

Chandrashekar Lakshminarayanan; Rahul Vashisht; Riya Mahesh

REVIEW 5 major objections 6 minor 23 references

Transformers with Sparse Attention for Granger Causality

T0 review · 5 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a transformer with sparse attention can recover Granger causal links in multivariate time series without a fixed lag, by learning which past time instances matter most.

desk verdict A clean, short sparse-attention Granger paper whose key lag-selection step is invalidated by the causal mask, and whose evaluation would not support the claims even without that flaw. read the letter →

arxiv 2411.13264 v1 pith:3AJO6SJ5 submitted 2024-11-20 cs.LG

Riya Mahesh , Rahul Vashisht , Chandrashekar Lakshminarayanan This is my paper · ORCID

classification cs.LG

keywords Multi-variatetimeseriesGrangerCausalityStationaritySelf-attentionSparseAttentionTransformersCausaldiscoveryVariablelags

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that a transformer whose attention mechanism is made sparse can recover Granger causality—whether past values of one series improve prediction of another—among variables in multivariate time series when the delays between cause and effect are small, unknown, and vary over time. The model first runs temporal attention over a sliding window, ranks past time instances by the column sums of the attention matrix, and keeps only the top k as the effective lag set; it then runs attention across variables and compares prediction errors with and without each variable masked. If the method works, causal discovery no longer needs a fixed lag specified in advance, which matters for real series where delays drift, such as climate or crop-yield data with shifting monsoon onsets. On synthetic time-homogeneous linear datasets with random delays, the paper reports that its Sparse Attention Transformer beats a fixed-lag vector autoregression (VAR) Granger baseline on AUC-ROC for all tested variable counts and on F1 for most.

What carries the argument

The carrying mechanism is the column-sum selection rule applied to the temporal attention matrix. In the temporal stage, self-attention is computed over a window of 2k past time steps with a causal mask; because each row of the resulting lower-triangular matrix sums to one, the sum of a column measures the average influence that one past time step exerts on all later tokens, and comparing column sums is equivalent to comparing column means. The paper takes the k columns with the largest sums as the model's learned choice of effective lags, and this selection is what allows the method to avoid a fixed lag. The second stage multiplies the selected steps' attention across variables and derives Granger indices by masking one variable column at a time during inference.

What would settle it

Generate a linear bivariate process where $X_t$ depends only on $Y_{t-10}$ plus small noise, train the Sparse Attention Transformer with window $2k = 20$ over many random seeds, and check the selected top-10 index set; if the true lag 10 is not consistently selected yet a VAR fit with the true lag detects the edge, the column-sum importance criterion has failed.

Watch

Extended reading notes

Core claim

The paper's central claim is that unknown, variable lags can be handled by a two-stage sparse attention mechanism rather than by fixing a lag in advance. Stage one treats the recent past as a window of 2k time tokens, applies a causal mask, and computes a temporal attention matrix; the column sums of this matrix are taken as importance scores, and the k columns with the largest sums select which past time instances form the effective input. Stage two transposes those k selected steps so the D variables become tokens, trains to predict the current vector, and at inference masks one variable at a time to obtain restricted predictions. The Granger indices are log ratios of restricted to unrestricted prediction error variances, normalized into a causation matrix. On synthetic linear, time-homogeneous datasets with maximum lag 10 (window 20, top k = 10), the paper reports that this model outperforms VAR-based Granger causality with the same window on AUC-ROC for D = 4, 5, 6, 10 and on F1 except D = 6.

Load-bearing premise

The claim rests on the unproven assumption that the column sums of the temporal attention matrix rank which past time instances matter for predicting the output; if that ranking is wrong, the selected lags and the resulting Granger indices lose their foundation.

Editorial extensions

If this is right

Practitioners can run Granger causal discovery on stationary multivariate series without choosing a lag in advance; the top-k selection supplies the effective past times contextually.
On series with random or shifting delays, the method reports higher AUC-ROC and F1 than fixed-lag VAR Granger causality on the synthetic benchmarks tested.
Because only k of the 2k window positions are retained and each variable is masked one at a time, the attention computation is sparse and does not require full pairwise attention across all time steps.
The output is a normalized D by D causation matrix comparable with ground-truth causal graphs and with classical Granger outputs, so existing evaluation metrics transfer directly.

Reading between the lines

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

If the column-sum interpretation is correct, the same top-k temporal selection could be used as a lag-selection preprocessor for standard Granger tests, giving classical methods an adaptive lag instead of a fixed one.
The paper leaves non-linear and non-stationary series to future work; a natural stress test is whether the column-sum ranking remains stable when the underlying causal strength changes over time, since attention statistics then shift.
The reported comparison uses the same window length for the VAR baseline; an oracle-lag VAR or a VAR with information-criterion lag selection would clarify how much of the gap comes from adaptive lag choice rather than from attention itself.
If the model genuinely learns context-dependent lags, the selected top-k index set should vary across windows when delays vary; measuring that variability would directly test whether the method is adapting rather than memorizing fixed positions.

Share X Bluesky LinkedIn Reddit HN

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

Colleague — read this one mainly as a cautionary example. The exact mechanism, top-k temporal lag selection via attention column sums followed by variable-masking attention for Granger indices, is new relative to the cited literature, and the paper is clearly written and easy to reproduce from the algorithm block. The synthetic C4C-style setup and the comparison against a fixed-lag VAR are appropriate for a first test. The citation pattern is fine; the related work section is short but relevant.

The load-bearing problem is in Section 4. The paper says column sums of the temporal attention matrix can be compared because they differ only by a scaling factor of 2k. That is false. The causal mask leaves only 2k-j+1 non-masked entries in column j. Even with uniform attention, column sums are H_{2k}-H_{j-1}, strictly decreasing in j, so the top-k selection always picks the oldest k lags. The stress-test note is correct. The claimed ability to "pick the most significant past time instances" is not demonstrated by the reported procedure, and the Granger indices computed after selection inherit the bias.

Secondary problems are real. The unrestricted-model errors appear to be computed on the training data, so the Granger indices may reflect overfitting rather than generalization. No standard deviations or significance tests are reported, and the VAR baseline with 20 lags on 150 steps is overparameterized. The experiment sets k=10 exactly equal to the known maximum lag, so it never tests the claimed advantage when the lag is unknown. No code is provided.

The idea is not without merit. A corrected version that normalizes column sums by the number of valid entries, evaluates on held-out data, and includes error bars would be worth another look. As written, the central novelty rests on a false normalization, and the evaluation is too weak to rescue it.

Who this is for: researchers interested in attention-based causal discovery, mainly as a demonstration of how easily attention-mask geometry can confound interpretability claims. I would not cite it in its current form. For peer review: I would have a serious referee look at it rather than desk-reject, because the flaw is precise and fixable, but I would not accept this version.

Referee Report

5 major / 6 minor

Summary. The paper proposes a Sparse Attention Transformer (SAT) for Granger causal discovery in multivariate time series with small, variable lags. The model first applies temporal self-attention over a 2k-length window with a causal mask, selects the k time instances with the largest column sums, and then applies inter-variable attention with leave-one-variable-out masking to produce Granger causality indices. The method is evaluated on synthetic linear datasets with up to 10 variables and 150 time steps, using AUC-ROC and F1 scores against a VAR-based Granger causality baseline. The paper claims that SAT significantly outperforms VAR and that the top-k temporal selection removes the need to specify a fixed lag manually.

Significance. If the central claims held, the paper would offer a practical attention-based alternative to fixed-lag VAR for Granger causal discovery, with a concrete algorithm and a falsifiable evaluation protocol on synthetic benchmarks. The two-stage attention design is clearly specified, and the authors provide a pseudocode algorithm and report results across four dataset groups. However, the main novelty — context-dependent selection of the most important past time instances — is undermined by a mask-induced bias in the column-sum comparison, and the empirical comparison with VAR is not statistically supported as written. The significance therefore depends on whether the temporal-selection mechanism can be corrected and the evaluation strengthened.

major comments (5)

[§4, temporal attention and Algorithm 1] The claim that comparing column sums is equivalent to comparing column means because they differ only by a scaling factor of 2k is false under the causal mask. In the temporal attention matrix A_t, entry (i,j) is nonzero only for i >= j, so column j contains 2k - j + 1 active entries, not 2k. For uniform attention with A_ij = 1/i, the column sum is H_{2k} - H_{j-1}, which strictly decreases with j; the top-k selection would then always choose the oldest k time steps. This count bias confounds the temporal selection and invalidates the paper's central claim that the model contextually picks the most significant past time instances. The authors should use a mask-aware normalization, such as averaging over the unmasked entries, or otherwise demonstrate that the selection is not driven by the mask.
[§6 and Table 1] The statement that SAT 'performs significantly better than the baseline method' is not supported statistically. No standard deviations, confidence intervals, or hypothesis tests are reported across the 20 datasets per group. In addition, the VAR baseline uses 20 lags on series of length 150 with up to 10 variables, which is severely overparameterized (roughly 20D coefficients per equation) and will underperform for reasons unrelated to Granger causal structure. A fair comparison requires a lag-selection procedure for VAR, such as BIC/AIC, or a lag length appropriate to the sample size, together with repeated-sampling variability measures.
[§4, Algorithm 1] The unrestricted-model errors appear to be computed on the training data: the pseudocode stores Y' from the backpropagation phase and the restricted-model predictions are also obtained on the same data. Training residuals can be made very small by overfitting, which makes the Granger index ln(sigma_R^2 / sigma_U^2) uninformative and potentially inflated. Evaluation of both unrestricted and restricted models should be performed on held-out data, and the authors should report the actual residual variances.
[§5.1] The setup sets k=10 because the synthetic data has a true maximum lag of 10 and the window size is 2k=20. This means the claimed advantage over 'manually feeding a fixed time lag value' is partly assumed rather than demonstrated. The model still requires the maximum lag / window size as a hyperparameter, and the experiments do not test robustness to misspecified k, lags larger than k, or variable lags with a wider range. At minimum, the paper should distinguish the known maximum lag used to construct the window from the unknown per-instance lags the model is supposed to select.
[§4, inter-variable attention and Granger masking] Masking a column of A_d with -infinity before softmax forces that variable's attention weight to zero in every row, but the paper does not establish that this operation is equivalent to Granger's restricted regression model. Attention weights are not shown to correspond to regression coefficients, and removing one variable also changes the softmax normalization for all other variables. As a concrete correctness check, I suggest a controlled experiment: when a variable that has no true causal effect is masked, the average CGCI for that variable should be near zero; if this does not hold, the proposed index measures model sensitivity rather than Granger causality.

minor comments (6)

[Table 1 caption] The caption says 'across the three groups of data samples' but the table reports four groups (D=4, 5, 6, 10); please correct the caption.
[§3.1] There is a typo: 'Autocorrelaion' should be 'Autocorrelation'.
[Algorithm 1] In the computation of s_l, the indices are inconsistent: s_l is written as a sum over j of A_{j,l}, but the loop variable and the later reference to 's_k values' should be aligned. Please rewrite the loop with clear row/column indices.
[§5.2] The positional embedding is mentioned only in the training details; the architecture description in §4 should state where positional embeddings are injected, since the temporal attention operates on the raw window.
[§6] The phrase 'significantly better' should be replaced by a quantitative statement backed by confidence intervals or statistical tests, as noted in the major comments.
[§1] The contribution statement says the model 'picks the top-k time-instances contextually,' but the subsequent description shows that k is a fixed hyperparameter equal to the true maximum lag in the experiments; the text should acknowledge this limitation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the temporal top-k masking bias is a correctness flaw, not a circular reduction.

full rationale

The derivation chain is not circular. The Granger causality output is defined independently of the model's attention weights: Algorithm 1 computes Â_lm = ln(σ²_lm / σ²_m) by comparing prediction-error variances under variable masking against the unrestricted model, and the method is evaluated against externally generated ground-truth causal graphs (Section 5.1), so the central causal claim is not defined in terms of the model's own fitted quantities. The attention-as-importance premise is borrowed from external prior work [8] and is explicitly left unproved in the conclusion ('Future work includes ... explaining the exact mathematical relation of multi-head attention weights with causality coefficients'), which is an assumption rather than a circular reduction. There are no load-bearing self-citations: every cited basis (Granger 1969, Vaswani et al. 2017, C4C challenge, etc.) is external. I do flag a serious non-circular correctness flaw in Section 4: with the causal mask M_{ij} = -∞ for i < j, column j has 2k−j+1 non-masked entries, so column sums and column means differ by a column-dependent factor, not the claimed constant 2k; under uniform attention s_j = H_{2k} − H_{j−1}, so the top-k selection is biased toward older lags. This undermines the temporal novelty claim but does not make the Granger causality output equivalent to the model's inputs by construction; the output is still a predictive-variance comparison against known ground truth.

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

The central claim rests on two unproven interpretive steps: attention weights measure causal influence, and masking attention inputs reproduces Granger's restricted regression. The only fitted choice is k=10, set to the known maximum lag, which weakens the novelty claim.

free parameters (3)

k (number of selected time steps) = 10
Set to the true maximum lag of the synthetic data (max lag 10). The paper's claim of not fixing the lag is only partial, because k is chosen from the known upper bound.
window size (2k) = 20
Derived from k=10; the model attends to a window of 20 past steps and keeps 10.
attention dimensions d_k1, d_k2 and training hyperparameters = not reported
The paper does not specify dimensions, learning rate, epochs, or batch size, so these are unstated free choices.

assumptions (4)

domain assumption Attention weights indicate causal links.
Adopted from Rohekar et al. (2024) and used as the foundation for interpreting attention matrices as causation, without proof in this paper.
ad hoc to paper Column sums of temporal attention identify the most important past time steps.
Key premise in Section 4 used for top-k selection; no theoretical or empirical validation independent of the paper's own framework.
ad hoc to paper Masking a variable's column in inter-variable attention reproduces Granger's restricted model.
Inference stage in Algorithm 1 assumes that ablating a variable in the attention layer is equivalent to removing it from the Granger regression, which is not established.
domain assumption Generation process is linear, time-homogeneous, with uniformly sampled random delays.
Borrowed from the C4C challenge framework; restricts the scope of the method and limits generalizability to nonlinear or non-stationary data.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Transformers with Sparse Attention for Granger Causality." pith.science (2026). https://pith.science/paper/3AJO6SJ5

@misc{pith2026241113264,
  author       = {Pith},
  title        = {Pith review of: Transformers with Sparse Attention for Granger Causality},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3AJO6SJ5}},
  note         = {Machine review of arXiv:2411.13264}
}

read the original abstract

Temporal causal analysis means understanding the underlying causes behind observed variables over time. Deep learning based methods such as transformers are increasingly used to capture temporal dynamics and causal relationships beyond mere correlations. Recent works suggest self-attention weights of transformers as a useful indicator of causal links. We leverage this to propose a novel modification to the self-attention module to establish causal links between the variables of multivariate time-series data with varying lag dependencies. Our Sparse Attention Transformer captures causal relationships using a two-fold approach - performing temporal attention first followed by attention between the variables across the time steps masking them individually to compute Granger Causality indices. The key novelty in our approach is the ability of the model to assert importance and pick the most significant past time instances for its prediction task against manually feeding a fixed time lag value. We demonstrate the effectiveness of our approach via extensive experimentation on several synthetic benchmark datasets. Furthermore, we compare the performance of our model with the traditional Vector Autoregression based Granger Causality method that assumes fixed lag length.

Figures

Figures reproduced from arXiv: 2411.13264 by the authors.

**Figure 1.** Sparse Attention Transformer Model Architecture · · · 𝑡 − 2𝑘 ≤ 𝑡 − 𝑘 − 𝑙1 ≤ 𝑡 − 𝑘 Thus, 𝑡 − 2𝑘 ≤ 𝑡 − 𝑖 − 𝑙𝑖 ≤ 𝑡 − 1 ∀ 𝑖 ∈ [1, 𝑘], which indicates the length of the sliding window to be 2𝑘 to establish the influence of past 2𝑘 time steps on the present values and pick the 𝑘 most important time instances. This is analogous to finding the 𝑘 unknown lags that determine the exact time steps causing the current vector. T… view at source ↗

**Figure 2.** ROC plots across four groups of data samples- D=4, 5, 6, 10; D indicating the number of variables in the time-series 6 RESULTS AND DISCUSSION We choose the AUC-ROC measure and F1 scores between the causation matrix and ground truth causality matrix as metrics to evaluate performance across various datasets, grouped into sets based on the number of variables in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 18 canonical work pages

[1]

Chainarong Amornbunchornvej, Elena Zheleva, and Tanya Y Berger-Wolf. 2019. Variable-lag granger causality for time series analysis. In 2019 IEEE International Conference on Data Science and Advanced Analytics (DSAA) . IEEE, 21–30

work page 2019
[2]

John Geweke. 1984. Inference and causality in economic time series models. Handbook of econometrics 2 (1984), 1101–1144

work page 1984
[3]

Clive WJ Granger. 1969. Investigating causal relations by econometric models and cross-spectral methods. Econometrica: journal of the Econometric Society (1969), 424–438

work page 1969
[4]

Uzma Hasan, Emam Hossain, and Md Osman Gani. 2023. A survey on causal discovery methods for iid and time series data. arXiv preprint arXiv:2303.15027 (2023)

arXiv 2023
[5]

Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic opti- mization. arXiv preprint arXiv:1412.6980 (2014)

arXiv 2014
[6]

Valentyn Melnychuk, Dennis Frauen, and Stefan Feuerriegel. 2022. Causal trans- former for estimating counterfactual outcomes. In International Conference on Machine Learning. PMLR, 15293–15329

work page 2022
[7]

Eshaan Nichani, Alex Damian, and Jason D Lee. 2024. How transformers learn causal structure with gradient descent. arXiv preprint arXiv:2402.14735 (2024)

arXiv 2024
[8]

Raanan Y Rohekar, Yaniv Gurwicz, and Shami Nisimov. 2024. Causal Inter- pretation of Self-Attention in Pre-Trained Transformers. Advances in Neural Information Processing Systems 36 (2024)

work page 2024

Show all 23 references

[9]

Jakob Runge, Sebastian Bathiany, Erik Bollt, Gustau Camps-Valls, Dim Coumou, Ethan Deyle, Clark Glymour, Marlene Kretschmer, Miguel D Mahecha, Jordi Muñoz-Marí, et al. 2019. Inferring causation from time series in Earth system sciences. Nature communications 10, 1 (2019), 2553

2019
[10]

Jakob Runge, Xavier-Andoni Tibau, Matthias Bruhns, Jordi Muñoz-Marí, and Gustau Camps-Valls. 2020. The causality for climate competition. In NeurIPS 2019 Competition and Demonstration Track . PMLR, 110–120

2020
[11]

Xiao Shou, Debarun Bhattacharjya, Tian Gao, Dharmashankar Subramanian, Oktie Hassanzadeh, and Kristin P Bennett. 2024. Pairwise causality guided transformers for event sequences. Advances in Neural Information Processing Systems 36 (2024)

2024
[12]

Elsa Siggiridou and Dimitris Kugiumtzis. 2015. Granger causality in multivariate time series using a time-ordered restricted vector autoregressive model. IEEE Transactions on Signal Processing 64, 7 (2015), 1759–1773

2015
[13]

Dmitry A Smirnov and Igor I Mokhov. 2009. From Granger causality to long-term causality: Application to climatic data. Physical Review E—Statistical, Nonlinear, and Soft Matter Physics 80, 1 (2009), 016208

2009
[14]

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. Advances in neural information processing systems 30 (2017)

2017
[15]

Sebastian Weichwald, Martin E Jakobsen, Phillip B Mogensen, Lasse Petersen, Nikolaj Thams, and Gherardo Varando. 2020. Causal structure learning from time series: Large regression coefficients may predict causal links better in practice than small p-values. InNeurIPS 2019 Comp...

2020
[16]

Dongxia Wu, Tsuyoshi Idé, Georgios Kollias, Jiri Navratil, Aurelie Lozano, Naoki Abe, Yian Ma, and Rose Yu. 2024. Learning Granger Causality from Instance- wise Self-attentive Hawkes Processes. In International Conference on Artificial Intelligence and Statistics. PMLR, 415–423

2024
[17]

Chonghao Zhang, Linlin Zhao, Zhefeng Yin, and Zhenguo Zhang. 2023. Causal- former: Causal Discovery-based Transformer for Multivariate Time Series Fore- casting. 1–6. https://doi.org/10.1109/CISP-BMEI60920.2023.10373365

2023
[18]

Cheng Zhang, Xiong Zou, and Chuan Lin. 2022. Fusing XGBoost and SHAP models for maritime accident prediction and causality interpretability analysis. Journal of Marine Science and Engineering 10, 8 (2022), 1154

2022
[19]

Wei Zhang, Thomas Panum, Somesh Jha, Prasad Chalasani, and David Page
[20]

Yifan Zhang, Rui Wu, Sergiu M Dascalu, and Frederick C Harris Jr. 2024. Sparse transformer with local and seasonal adaptation for multivariate time series fore- casting. Scientific Reports 14, 1 (2024), 15909

2024
[21]

Haoyi Zhou, Shanghang Zhang, Jieqi Peng, Shuai Zhang, Jianxin Li, Hui Xiong, and Wancai Zhang. 2021. Informer: Beyond efficient transformer for long se- quence time-series forecasting. In Proceedings of the AAAI conference on artificial intelligence, Vol. 35. 11106–11115

2021
[22]

Yaqi Zhu, Fan Yang, and Andrei Torgashov. 2024. Causal-Transformer: Spatial- temporal causal attention-based transformer for time series prediction. IFAC- PapersOnLine 58, 14 (2024), 79–84

2024
[2020]

In International Conference on Machine Learning

Cause: Learning granger causality from event sequences using attribution methods. In International Conference on Machine Learning . PMLR, 11235–11245

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Chainarong Amornbunchornvej, Elena Zheleva, and Tanya Y Berger-Wolf. 2019. Variable-lag granger causality for time series analysis. In 2019 IEEE International Conference on Data Science and Advanced Analytics (DSAA) . IEEE, 21–30

work page 2019

[2] [2]

John Geweke. 1984. Inference and causality in economic time series models. Handbook of econometrics 2 (1984), 1101–1144

work page 1984

[3] [3]

Clive WJ Granger. 1969. Investigating causal relations by econometric models and cross-spectral methods. Econometrica: journal of the Econometric Society (1969), 424–438

work page 1969

[4] [4]

Uzma Hasan, Emam Hossain, and Md Osman Gani. 2023. A survey on causal discovery methods for iid and time series data. arXiv preprint arXiv:2303.15027 (2023)

arXiv 2023

[5] [5]

Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic opti- mization. arXiv preprint arXiv:1412.6980 (2014)

arXiv 2014

[6] [6]

Valentyn Melnychuk, Dennis Frauen, and Stefan Feuerriegel. 2022. Causal trans- former for estimating counterfactual outcomes. In International Conference on Machine Learning. PMLR, 15293–15329

work page 2022

[7] [7]

Eshaan Nichani, Alex Damian, and Jason D Lee. 2024. How transformers learn causal structure with gradient descent. arXiv preprint arXiv:2402.14735 (2024)

arXiv 2024

[8] [8]

Raanan Y Rohekar, Yaniv Gurwicz, and Shami Nisimov. 2024. Causal Inter- pretation of Self-Attention in Pre-Trained Transformers. Advances in Neural Information Processing Systems 36 (2024)

work page 2024