acceptodds
Under review as a conference paper at ICLR 2027

A Single Leaf Heralds Autumn: Predicting Battery Lifetime under Stationarity Masking

Abstract

Early battery lifetime prediction estimates long-term cycle life from limited initial cycles, reducing the time required for full degradation testing. However, early cycles are highly similar, making subtle degradation variations difficult to identify from individual cycle states. Beyond marginal distribution shifts in time-series non-stationarity, we characterize a regime in which near-stationary marginals coexist with variations in conditional transitions between adjacent cycles, which we term stationarity masking. To isolate these hidden transition variations from marginal effects, we disentangle marginal and conditional discrepancies and derive a unique transition projection that preserves the marginal distributions of both adjacent cycles while capturing their inter-cycle transition structure. Building on this analysis, we propose LEAF, a simple yet effective framework for early battery lifetime prediction. LEAF learns conditional associations between local states of adjacent cycles, extracts conditional transitions through fixed-marginal projection, and models their sequential evolution for lifetime prediction. Extensive experiments on 16 datasets spanning Li-ion, Zn-ion, and Na-ion batteries validate the effectiveness of LEAF across diverse battery systems and aging conditions. Code is available at https://anonymous.4open.science/r/Batt_LEAF/.

Primary area:
learning on time series and dynamical systems
Keywords:
Battery Lifetime Prediction, Stationarity Masking, Early Degradation, Time Series Modeling

Reviews

Among:
Shape:
1234this paper, Bottom 10%
Figure 1. Mean review rating of the reviewed papers at ICLR 2027. This paper averages 1.67 over 3 reviews, above 10% of the others.
Reviewer_XmJUOpenReview ↗
Rating
2
Confidence
2
Informativeness
2
Presentation
2
Soundness
2
Read review
Paper summary

The paper studies early battery lifetime prediction, where the goal is to estimate a battery’s eventual cycle life from a limited number of initial charge–discharge cycles. Its central motivation is that adjacent early cycles can have very similar marginal distributions even when their cross-cycle dependencies evolve with degradation. The authors call this phenomenon stationarity masking. The paper proposes LEAF, a framework that models transitions between local states of adjacent battery cycles rather than relying exclusively on individual-cycle representations.

The theoretical motivation combines the KL chain rule, which separates marginal and conditional discrepancies, with a projection onto a fixed-marginal feasible set. The authors evaluate LEAF on 16 battery datasets spanning Li-ion, Na-ion, and Zn-ion batteries, comparing it with eight baselines. They report improvements in mean absolute percentage error (MAPE) and 15%-Acc across four battery domains, alongside ablation, transferability, and computational-efficiency studies.

Critical strengths

S1: A useful perspective on weak early-stage degradation signals

The paper identifies a plausible limitation of marginal-distribution-based descriptions: similar distributions of individual cycles do not imply similar relationships between consecutive cycles. Shifting attention toward conditional transitions is a natural and potentially useful way to capture subtle temporal changes. The KL decomposition provides a clear statistical framing for distinguishing marginal differences from conditional differences. This gives the proposed method a coherent motivation rather than introducing an architecture without a well-defined problem.

S2: Broad empirical evaluation with component ablations

The evaluation covers 16 datasets organized into four battery domains and compares LEAF with eight methods, including general time-series models and battery-specific approaches. The reported domain-level MAPE improves over the strongest baseline by 13.0% for Li-ion, 18.7% for CALB, 19.5% for Na-ion, and 16.3% for Zn-ion. The ablations also show consistent degradation when removing conditional transitions, fixed-marginal projection, transition encoding, or temporal order. These results provide encouraging evidence that the overall design is effective and that its components may contribute complementary information.

S3: A lightweight, interpretable representation

The fixed-marginal construction explicitly separates state composition from cross-cycle dependence. Its resulting transition matrices can be inspected to identify which local state associations are enhanced or suppressed relative to independence. The paper also reports favorable computational costs compared with several more complex baselines. This combination of a relatively simple architecture, inspectable intermediate representations, and competitive predictive performance is a practical strength.

Critical weaknesses

W1: The methodological novelty may be more limited than the theoretical presentation suggests

The KL chain rule and the uniqueness of Euclidean projection onto a nonempty closed convex set are standard mathematical results. Their application here is reasonable, but the paper needs to establish more clearly what is technically new beyond applying these results to adjacent-cycle battery representations. In particular, the fixed-marginal projection guarantees a feasible joint distribution with the specified marginals, but this property alone does not establish that the resulting representation is more informative about degradation or more physically meaningful. The empirical improvements are encouraging, yet the claimed advantage of this particular projection over other ways of normalizing or correcting transitions needs stronger isolation. The key issue is not whether the mathematics is correct, but whether the proposed construction constitutes a sufficiently distinctive methodological contribution for ICLR.

W2: The evidence for “stationarity masking” and its causal role in the gains is incomplete

The paper shows examples in which conditional-transition discrepancies exceed marginal discrepancies. However, these observations do not by themselves establish that stationarity masking is widespread, that the discrepancy reflects degradation rather than noise or operating-condition variation, or that exploiting it explains the predictive improvements. The analysis uses learned local states and conditional representations, so the observed discrepancies may depend on the encoder and its training objective. A stronger demonstration would compare marginal and conditional discrepancies under controlled conditions, quantify uncertainty across batteries, and establish that conditional information adds predictive value beyond comparable representations and model capacity. The ablations support the utility of the components, but they do not fully validate the proposed explanation for why the method works.

W3: Generalization and statistical evidence need more careful qualification

The main results are promising, but the reported standard deviations come from three runs, and the manuscript does not foreground paired significance tests or confidence intervals for the improvements over the strongest baselines. This matters particularly for the smaller datasets, including CALB and Na-ion. The transfer results also show that direct transfer of the pretrained model performs poorly, while adaptation is needed; domain adaptation does not outperform the target-only baseline on CALB. Thus, the results support the usefulness of LEAF under the evaluated settings more strongly than they establish broad transferability without substantial target-domain supervision. The authors should clarify the scope of the generalization claims and provide stronger uncertainty estimates or targeted tests on genuinely unseen conditions.

Questions for authors

Q1. What is the incremental contribution of fixed-marginal projection?

The Euclidean projection preserves the two cycle marginals while adjusting the joint transition matrix. How does LEAF compare with alternative ways of obtaining a joint distribution with fixed marginals, such as entropy-regularized transport, iterative proportional fitting, or simpler normalization schemes? Can the authors show that the particular Euclidean projection is important for prediction rather than merely enforcing marginal consistency?

Q2. How robust is the evidence for stationarity masking?

The paper reports larger conditional than marginal discrepancies across multiple datasets. Are these results robust across random battery selections, different patch counts, and different encoders? Could the authors provide uncertainty estimates and a controlled comparison demonstrating that the conditional discrepancies contain degradation-related information beyond measurement noise and operating-condition differences?

Q3. How statistically reliable are the reported improvements?

Given that results are averaged over three runs, can the authors provide paired comparisons against the strongest baselines, with confidence intervals or appropriate statistical tests? Since multiple prefixes from the same battery share a lifetime label, how sensitive are the results to evaluating and aggregating performance at the battery level rather than treating all prefix predictions as separate evaluation samples?

Q4. Under what conditions does LEAF generalize to unseen batteries and domains?

The transfer experiments show that direct transfer performs poorly and that adaptation is important. Can the authors clarify how much labeled target-domain data is required to obtain the reported gains, and whether LEAF consistently outperforms target-only training under strictly matched data budgets across all target domains? How well does it perform when the target chemistry or operating protocol is genuinely unseen?

1 comment

As there were no responses I will keep my scores.

Reviewer_qMJROpenReview ↗
Rating
1
Confidence
3
Informativeness
2
Presentation
2
Soundness
1
Read review
Paper summary

This paper addresses the problem of predicting battery lifetime from a sequence of early, controlled charge–discharge cycles. Since the waveform of each cycle is dominated by normal behavior by design, simply using the time-series data as input features carries a significant risk of diluting subtle but critical deviations that may affect battery life. Motivated by this, the authors focus on the transition probability between two neighboring cycles.

The authors provide two theorems on basic properties of joint distributions and their marginals. However, the main text lacks a self-contained description of the proposed regression model, which makes it difficult to assess the novelty.

Critical strengths

The problem and its unique challenges are clearly described in Sections 1 and 2.1. In particular, the point that lifetime prediction requires capturing subtle changes in the observed waveforms over cycles is well articulated.

Critical weaknesses
  • The concept of stationarity masking (Definition 3.1) amounts to the standard distinction between marginals and conditionals. For any process with dependence between adjacent cycles, conditionals can vary while marginals stay nearly fixed. It is unclear whether the authors claim novelty in this concept.
  • Theorem 3.1 is the chain rule of the KL divergence. It is not novel, and it has no organic connection to the proposed model, which neither computes nor optimizes any KL divergence.
  • Theorem 3.2 is the well-known existence and uniqueness of the Euclidean projection onto a nonempty closed convex set. It only guarantees that the projection in Eq. (11) is well defined; it does not explain why this projection should help lifetime prediction.
  • The model described in Section 3.2 does not appear to implement the motivation in the Introduction. Moreover, nothing in the model appears to address the issue that waveforms are dominated by normal behavior, which the Introduction identifies as the key challenge.
  • The main text does not provide a self-contained definition of the regression function fθf_\theta.
Questions for authors
  1. Please provide a self-contained description of how the research question articulated in the Introduction is addressed by the proposed regression model.
  2. In what sense do PcP_c and KcK_c, defined over patch indices, represent the distributions in Definition 3.1, which are defined over cycle observations?
Reviewer_qkcnOpenReview ↗
Rating
2
Confidence
3
Informativeness
2
Presentation
3
Soundness
2
Read review
Paper summary

LEAF addresses early battery lifetime prediction, where early cycles are highly similar and changes in the relationships between cycles may be hidden by nearly unchanged marginal distributions. It learns local associations between adjacent cycles, uses a fixed-marginal projection to build consistent transition representations, and models their changes over time to predict battery lifetime. Experiments on 16 battery datasets show that LEAF outperforms 8 baseline methods, reducing MAPE compared with the strongest baseline in each battery domain.

Critical strengths
  1. Clear problem statement. The paper clearly describes the research problem and provides a well-motivated approach.
  2. Code availability. The authors provide code, which supports reproducibility and makes it easier for others to build on this work.
  3. Strong experimental results. LEAF outperforms 8 baselines across 16 battery datasets, showing consistent improvements across different battery types.
Critical weaknesses
  1. Unclear novelty and incomplete related work. The proposed method uses marginal-constrained soft matching and Euclidean projection to model cross-cycle transitions. The related work should better discuss these mechanisms, including closely related studies [1, 2] and more recent methods. The authors should clarify the main contribution beyond adapting these ideas to battery lifetime prediction.
  2. Unclear experimental settings and insufficient evidence. Some baseline results differ substantially from those reported in BatteryLife [3]. The authors should clarify the experimental settings. Moreover, the experiments do not directly show that stationarity masking limits existing methods.

[1] Liu, Y., Zhu, L., Yamada, M., and Yang, Y. (2020). Semantic Correspondence as an Optimal Transport Problem. [2] Blondel, M., Seguy, V., and Rolet, A. (2018). Smooth and Sparse Optimal Transport. [3] Tan, R., et al. (2025). BatteryLife: A Comprehensive Dataset and Benchmark for Battery Life Prediction.

Questions for authors
  1. Could the authors clarify the main new contribution of LEAF compared with prior studies [1–2] and battery lifetime prediction methods such as DITING? In particular, what new insight or battery-specific design does LEAF provide?
  2. Several baseline results in Sections 4.1–4.2 and Table 1 differ substantially from those reported in BatteryLife [3]. Did the authors use the same dataset versions, lifetime labels, data splits, preprocessing, and tuning settings? Could the authors explain these differences?
  3. In the ablation study (Section 4.3, Table 2), without fixed-marginal projection (w/o FP), LEAF still outperforms the strongest baselines in three domains. Even with shuffled transition order (w/o TO), LEAF still performs better in all four domains. What explains these results?
  4. Regarding the evidence for stationarity masking in Section 4.4 (Q1) and Figure 3, how are the marginal and conditional distributions estimated? How do these results support the existence of stationarity masking in battery cycles?

From OpenReview, CC BY 4.0. Authors as signed there; shown unedited.

Related papers

Loading the map…

Discussion (0)

Sign in to comment.