Pith. sign in

REVIEW 3 major objections 7 minor 25 references

AnomSeer claims a 3B vision-language model can beat GPT-4o at time-series anomaly detection by grounding its reasoning in classical statistics via a new RL objective.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-03 03:07 UTC pith:73JUYNBO

load-bearing objection The detection gains are probably real, but the paper's reasoning story leaks the answer through ExpCoT, so the causal claim is unproven. the 3 major comments →

arxiv 2602.08868 v2 pith:73JUYNBO submitted 2026-02-09 cs.LG cs.AI

AnomSeer: Reinforcing Multimodal LLMs to Reason for Time-Series Anomaly Detection

Junru Zhang , Lang Feng , Haoran Shi , Xu Guo , Han Yu , Yabo Dong , Duanqing Xu This is my paper
classification cs.LG cs.AI
keywords time-series anomaly detectionmultimodal LLMreinforcement learningchain-of-thoughtoptimal transportpolicy optimizationgeneralizationreasoning grounding
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper tries to establish that multimodal LLMs' weakness in time-series anomaly detection—relying on coarse visual heuristics—can be fixed by a reinforcement-learning post-training method that aligns the model's reasoning with expert statistical traces. It introduces AnomSeer, built on ExpCoT (expert chain-of-thought) and TimerPO (time-series grounded policy optimization), and claims that a 3B/7B open model outperforms far larger commercial models such as GPT-4o in both classification and localization while producing verifiable reasoning traces. If correct, this would mean small open models can unify anomaly detection, type classification, and explanation, and generalize to unseen anomaly types and real-world benchmarks. The load-bearing premise is that the expert traces are generated from ground-truth annotations, so the causal role of fine-grained reasoning must be tested against the possibility of post-hoc rationalization.

Core claim

On its own terms, the paper discovers that a purely RL-trained MLLM can be steered away from coarse 'eyeballing' toward fine-grained, numerically grounded reasoning by rewarding semantic proximity to expert-designed reasoning traces. Using a 3B-parameter Qwen2.5-VL backbone, it achieves 62.8% classification accuracy and 79.3% Affinity-F1 on AnomLLM, versus GPT-4o's 17.2% and 41.2%, with larger gains on frequency- and point-driven anomalies; it also detects shapelet anomalies that were absent from training and sustains performance on the real-world TSB-UAD benchmark. The central claim is that the orthogonalized optimal-transport reasoning advantage, not merely outcome rewards, is what instill

What carries the argument

ExpCoT (expert chain-of-thought): a three-stage reasoning trace (Observation → Reasoning & Validation → Conclusion) generated per instance from ground-truth annotations and classical TSAD analyses (histogram outlier scores, FFT periodograms, matrix profile discords, smoothed-gradient checks). TimerPO (time-series grounded policy optimization): a GRPO-style RL objective that adds a time-series reasoning advantage computed as the entropic-regularized optimal transport distance between model reasoning tokens and ExpCoT tokens, then orthogonalizes that advantage against the main detection advantage before combining, so the auxiliary signal refines reasoning without interfering with the primary o

Load-bearing premise

The expert reasoning traces are generated from ground-truth anomaly annotations, so the reward may be teaching the model to produce plausible-sounding justifications for known labels rather than to perform independent fine-grained analysis.

What would settle it

Generate ExpCoT traces while withholding the ground-truth interval from the Conclusion stage (or from the Reasoning & Validation selection), then retrain AnomSeer and evaluate on AnomLLM test: if classification/localization accuracy collapses to near-baseline levels, the fine-grained reasoning trace is not causally responsible for the gains; conversely, if it holds up, the model has internalized the statistical validation rather than the GT answer.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Unifying detection, categorization, and explanation in a single generative model becomes achievable for small open models, not just large proprietary ones.
  • Training on synthetic anomalies can transfer to unseen anomaly categories (e.g., shapelets) and real-world datasets, suggesting RL-from-classical-priors may generalize beyond memorization.
  • The orthogonal-projection trick offers a template for injecting auxiliary dense reward signals into RL without corrupting the primary task objective.
  • If verified, the approach could make anomaly explanations auditable, since they are tied to quantitative statistical evidence rather than free-form rationalization.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Since ExpCoT is built from the ground-truth interval, the reasoning advantage may be teaching the model to cite evidence that supports a known answer rather than to discover anomalies de novo; a clean test is to omit the GT interval from the Conclusion stage during trace generation and measure the drop.
  • The method's dependence on classical TSAD primitives (e.g., matrix profile, FFT) suggests a natural ceiling: if those primitives are blind to certain anomalies, the ExpCoT supervision will also be blind, so the model's generalization may be bounded by the classical toolkit's coverage.
  • A testable extension is to apply the same TimerPO recipe to multivariate series by rendering each channel as a sub-image, as the paper itself suggests; the orthogonalization should mitigate cross-channel interference in the same way it mitigates task/reasoning interference.
  • The reasoning-trace distributions (token usage shift from 'global/sudden' to 'timestamp/amplitude') could serve as a probe for whether the model is truly grounded: an adversarial input with mismatched statistics should break the alignment if the reasoning is causal rather than rhetorical.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 7 minor

Summary. The paper proposes AnomSeer, a reinforcement-learning post-training method for multimodal LLMs (MLLMs) to perform time-series anomaly detection (TSAD) by unifying anomaly classification, localization, and explanation. AnomSeer builds on Qwen2.5-VL-3B/7B and introduces two components: (i) an expert chain-of-thought (ExpCoT) trace generated from classical TSAD analyses, and (ii) TimerPO, an RL algorithm that uses an optimal-transport-based reasoning reward and an orthogonal projection to combine this reward with the outcome-based advantage. The authors train on 3,200 synthetic AnomLLM instances and evaluate on AnomLLM, VisualTimeAnomaly, and TSB-UAD, reporting that AnomSeer-3B outperforms large commercial MLLMs such as GPT-4o and the prior RL-based TimeMaster in classification accuracy and Affinity-F1, while also producing structured, plausibly grounded reasoning traces. The central claim is that the fine-grained statistical reasoning guided by ExpCoT is the cause of the detection gains.

Significance. If the reported results hold, the paper would demonstrate that a lightweight open-source MLLM can outperform much larger commercial models on TSAD and generalize to unseen anomaly types, which is a practically valuable result for resource-constrained deployment. The paper's strengths include evaluation on three diverse benchmarks, a clear ablation study showing each component contributes, and the inclusion of pseudocode and hyperparameter details. However, the significance is tempered by two concerns: the ExpCoT traces are explicitly generated from ground-truth anomaly labels, so the reasoning reward may teach answer-aware rationalization rather than independent statistical analysis; and the reasoning evaluation in Section 5.3 measures alignment to the very same ExpCoT used for training, making the interpretability evidence circular. The method is nonetheless a substantive empirical contribution to RL for time-series reasoning, but the causal claim about reasoning driving detection needs stronger support.

major comments (3)
  1. [§4.2, Eq. (6)-(8)] ExpCoT is generated per instance from the ground-truth class and temporal annotation (Appendix B.2 explicitly states that 'REASONING & VALIDATION selects the analysis matched to the GT type' and 'CONCLUSION retains the GT interval(s) as the final localization'). The reasoning reward in Eq. (5) is the OT distance between the model response and this answer-embedded trace. Consequently, TimerPO rewards the model for producing text that resembles a post-hoc rationalization of the known label, not necessarily for deriving evidence from the plot. The ablation in Table 2 shows that removing ExpCoT degrades performance, but this does not distinguish 'grounded statistical reasoning' from 'distillation of the answer through the trace.' Section 5.3 and Figure 5 evaluate distributional alignment to the same ExpCoT used as the training target, so the reasoning evaluation is circular. I request a GT-f
  2. [§5.4, Figure 6] The orthogonal projection in Eq. (6) is applied to the group-level advantage vectors bAmain and bATsR in R^G, while the policy gradient in Eq. (8) is computed per token using importance ratios ρ_n^i. Orthogonality of the group-level advantage vectors does not imply that the resulting parameter gradients are orthogonal or that the auxiliary signal 'does not interfere with the primary detection objective.' The paper should either provide a mathematical argument for gradient-level non-interference or present empirical evidence (e.g., cosine similarity between the gradients from each objective during training) to substantiate the claim that orthogonalization prevents interference.
  3. Generalization results in Figure 6 are presented as unaveraged bar charts without error bars or significance testing across the 1,635 TSB-UAD series. This makes it difficult to assess whether the reported gains over baselines are consistent across domains or driven by a few favorable series. Please report per-dataset distribution statistics (e.g., mean ± std, box plots) and a paired significance test (e.g., Wilcoxon signed-rank or paired bootstrap) across the collection.
minor comments (7)
  1. [Abstract/§4.1] The abstract describes ExpCoT as 'verifiable, fine-grained reasoning,' but since the trace is generated from GT annotations, it is not independently verifiable. Please rephrase to avoid overclaiming.
  2. [Reproducibility] No code or data are released. Given the many hyperparameters and the threshold choices in ExpCoT (e.g., z>3.5, m=50), a public repository is needed for reproduction.
  3. [Figure 6] The y-axis begins at 18, which visually exaggerates differences between models. Consider starting at 0 and adding error bars.
  4. [Table 1] The table layout is dense and hard to parse. Recommend splitting into separate tables for classification and localization metrics, or using grouped sub-headers.
  5. [§5.3, Figure 5] The token-usage plots are qualitative. Report quantitative similarity measures (e.g., TF-IDF cosine similarity or Jaccard overlap) between ExpCoT and model outputs to support the distributional-alignment claim.
  6. [Appendix C.3, Table 4] Several hyperparameters are missing: the temperature τ in Eq. (5), the Sinkhorn regularization parameter, and the ExpCoT thresholds (z>3.5, window m=50, gradient window, FFT MAD tolerance). Please include these for reproducibility.
  7. [§4.2] The text says 'pure RL training strategy ... without SFT as a cold-start,' but the model is initialized from Qwen2.5-VL-Instruct, which is already instruction-tuned. Clarify that this is RL post-training on top of an instruction-tuned model.

Circularity Check

2 steps flagged

ExpCoT is built from ground-truth labels and retains the GT interval as its conclusion, so TimerPO's reasoning reward is answer-aware; the reasoning-quality evaluation then measures alignment to the same ExpCoT used as the training target, making the 'grounded reasoning' evidence circular even though detection is benchmarked externally.

specific steps
  1. fitted input called prediction [Appendix B.2, 'Instantiation with Ground Truth'; Eqs. (4)-(7), Section 4.2]
    "For every instance, ExpCoT is generated from the GT class and temporal annotation: OBSERVATION anchors on the GT interval(s) and applies the unified scan ... REASONING & VALIDATION then selects the analysis matched to the GT type and reports concrete numerical evidence ... CONCLUSION integrates these results and retains the GT interval(s) as the final localization, yielding a faithful, interpretable trace for supervising MLLM training."

    ExpCoT, the supervision target of TimerPO, is constructed from the ground-truth answer and its Conclusion explicitly retains the GT interval(s). The OT distance in Eq. (4) and the reasoning advantage in Eq. (5) are computed against this answer-containing trace, so the RL reward trains the model to reproduce a post-hoc rationalization of the known label. The orthogonal projection in Eq. (6) only decorrelates the reasoning advantage from the outcome advantage in group-vector space; it does not remove the GT answer from the target trace. The paper does not test whether the same statistical analysis would be generated without knowledge of the GT annotation.

  2. self definitional [Section 5.3, Figure 5; Appendix D.1, Figure 12]
    "To show that ANOMSEER enables time-series MLLMs reasoning grounded in fine-grained statistics, we analyze the effect of TimerPO on distributional alignment and linguistic usage before and after RL training... After TimerPO (d), the vocabulary shifts toward finer-grained and temporally grounded tokens (e.g., timestamp, intervals, amplitude), which better capture structured reasoning over time. Therefore, these results demonstrate that TimerPO not only improves distributional alignment with expert reasoning but also enriches the semantic granularity of reasoning traces, moving from broad anomaly"

    The evaluation uses the same ExpCoT that Eqs. (4)-(7) use as the optimization target. The OT distance in Eq. (4) measures semantic deviation from ExpCoT and Eq. (5) converts it into the reasoning advantage, so improved alignment after training is a restatement of optimizing the training reward, not an independent measurement of whether the model's reasoning is actually grounded in classical statistical analysis. No blind condition is provided in which ExpCoT does not contain the GT answer, and no test is run to determine whether the reasoning traces would arise from the statistical methods without prior knowledge of the labels.

full rationale

The central detection claims (classification accuracy and Affinity-F1 on the AnomLLM test set, VisualTimeAnomaly, and TSB-UAD) are not circular: AnomSeer is trained only on 3,200 synthetic AnomLLM instances and evaluated on held-out and external benchmarks, so those numbers provide independent support. The circularity is confined to the reasoning-quality claim. ExpCoT—the target of the OT-based reasoning advantage—is generated from the GT class and temporal annotation, and its Conclusion stage expressly retains the GT interval(s). Thus the RL reward rewards responses that resemble a post-hoc rationalization of the known answer. Section 5.3 and Appendix D.1 then use distributional alignment to that same ExpCoT as evidence that TimerPO produces 'grounded' reasoning; this is a report on the training objective, not an independent test. The orthogonal projection in Eq. (6) only decorrelates the two advantage vectors; it does not remove the answer from the trace, nor does the paper test whether the statistical analysis would be produced without GT. These issues make the reasoning-improvement evidence partially self-referential, but they do not undermine the externally benchmarked detection result. No load-bearing self-citations or imported uniqueness theorems were found.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The central claim depends on hand-set reward weights, the alpha weight, an unspecified temperature, and a battery of thresholds used to generate ExpCoT supervision. The training also assumes that plotted images retain sufficient information and that classical TSAD statistics anchored to ground truth form valid reasoning targets. No new physical entities are introduced.

free parameters (4)
  • lambda_fmt, lambda_cls, lambda_loc = 0.1, 0.2, 0.7
    Outcome reward weights, empirically chosen (Table 4).
  • alpha = 0.3
    Reasoning advantage weight; selected via sensitivity analysis on AnomLLM (Figure 4), though robust in [0.3,0.7].
  • tau (temperature)
    Appears in r_TsR = exp(-W/tau) (Eq. 5); value not reported, affecting the scale of reasoning rewards.
  • ExpCoT thresholds (z>3.5, window m=50, gradient k, FFT MAD tolerance) = z=3.5, m=50, etc.
    Hand-set statistical thresholds in ExpCoT generation (Appendix B.2); these shape the supervision traces and are not fitted to external data.
axioms (4)
  • domain assumption Visual line-plot rendering preserves enough information for MLLMs to detect anomalies.
    Section 3 follows the visualization input strategy; if plots discard key numerical detail, the whole approach is compromised.
  • domain assumption Classical TSAD heuristics (HBOS, Matrix Profile, FFT periodogram, gradient analysis) provide valid evidence when anchored to ground-truth intervals.
    ExpCoT generation (Section 4.1) assumes these methods' outputs justify the GT labels.
  • ad hoc to paper ExpCoT traces generated from ground truth are appropriate RL guidance.
    Appendix B.2: 'ExpCoT is instantiated per instance from the ground-truth anomaly type and temporal annotation' and 'retains the GT interval(s) as the definitive localization'.
  • ad hoc to paper The OT reasoning advantage based on the policy's own token embeddings and the orthogonal projection produce non-interfering, semantically meaningful guidance.
    Section 4.2: embeddings e_i from the policy itself define cost matrix C_i; orthogonality assumption in Eq. (6) is a design postulate without independent validation.

pith-pipeline@v1.3.0-alltime-deepseek · 19687 in / 11450 out tokens · 115761 ms · 2026-08-03T03:07:11.084569+00:00 · methodology

0 comments
read the original abstract

Time-series anomaly detection (TSAD) with multimodal large language models (MLLMs) is an emerging area, yet a persistent challenge remains: MLLMs rely on coarse time-series heuristics but struggle with multi-dimensional, detailed reasoning, which is vital for understanding complex time-series data. We present AnomSeer to address this by reinforcing the model to ground its reasoning in precise, structural details of time series, unifying anomaly classification, localization, and explanation. At its core, an expert chain-of-thought trace is generated to provide a verifiable, fine-grained reasoning from classical analyses (e.g., statistical measures, frequency transforms). Building on this, we propose a novel time-series grounded policy optimization (TimerPO) that incorporates two additional components beyond standard reinforcement learning: a time-series grounded advantage based on optimal transport and an orthogonal projection to ensure this auxiliary granular signal does not interfere with the primary detection objective. Across diverse anomaly scenarios, AnomSeer, with Qwen2.5-VL-3B/7B-Instruct, outperforms larger commercial baselines (e.g., GPT-4o) in classification and localization accuracy, particularly on point- and frequency-driven exceptions. Moreover, it produces plausible time-series reasoning traces that support its conclusions.

Figures

Figures reproduced from arXiv: 2602.08868 by Duanqing Xu, Han Yu, Haoran Shi, Junru Zhang, Lang Feng, Xu Guo, Yabo Dong.

Figure 1
Figure 1. Figure 1: Comparison of model performance and time-series reasoning quality. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The overall framework of ANOMSEER. ANOMSEER first generates ExpCoT reasoning traces y ∗ from the time-series data based on classical TSAD techniques (e.g., FFT). TimerPO then computes the outcome-aware advantage and leverages optimal transport to compute the time-series reasoning advantage, which is orthogonally integrated into policy optimization to ensure stable train￾ing and improved reasoning quality. … view at source ↗
Figure 3
Figure 3. Figure 3: An example of TSAD reason￾ing produced by ANOMSEER. Overall. ANOMSEER employs the pure RL training strat￾egy to enhance MLLMs without SFT as a cold-start or any modifications to the model architecture. During train￾ing, we first construct ExpCoT using the analytical rigor of traditional TSAD methods, and subsequently refine the model’s policy using orthogonalized time-series reason￾ing advantages through T… view at source ↗
Figure 4
Figure 4. Figure 4: Hyperparameter sensitivity analysis on α, comparing our method with the GPT-4o baseline (grey dashed line) [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of distribution alignment between ExpCoT (blue) and A [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Comparison of model generaliza￾tion performance (Affinity F1%) across point￾wise tasks, range-wise tasks, and the real￾world TSB-UAD benchmark. At last, we evaluate the generalization ability of ANOMSEER. We test the model (trained on the synthetic AnomLLM) on two distinct and more challenging benchmarks: VisualTimeAnomaly (a hybrid synthetic-real dataset with richer anomaly types) and TSB-UAD (a real-worl… view at source ↗
Figure 7
Figure 7. Figure 7: Prompt definition for time-series anomaly detection [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Example of ExpCoT reasoning trace for contextual point anomaly detection. [PITH_FULL_IMAGE:figures/full_fig_p015_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: ExpCoT reasoning trace for global point (out-of-range) anomaly detection. [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: ExpCoT reasoning trace for trend shift anomaly detection. [PITH_FULL_IMAGE:figures/full_fig_p016_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: ExpCoT reasoning trace for frequency deviation anomaly detection. [PITH_FULL_IMAGE:figures/full_fig_p017_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Comparison of distributional alignment between ExpCoT (blue) and A [PITH_FULL_IMAGE:figures/full_fig_p018_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Failure case 21 [PITH_FULL_IMAGE:figures/full_fig_p021_13.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

25 extracted references · 16 linked inside Pith

  1. [1]

    Large language models can be zero-shot anomaly detectors for time series?arXiv preprint arXiv:2405.14755,

    Sarah Alnegheimish, Linh Nguyen, Laure Berti-Equille, and Kalyan Veeramachaneni. Large language models can be zero-shot anomaly detectors for time series?arXiv preprint arXiv:2405.14755,

    Pith/arXiv arXiv

  2. [5]

    Histogram-based outlier score (HBOS): A fast unsupervised anomaly detection algorithm.KI-2012: poster and demo track, 1:59–63,

    Markus Goldstein and Andreas Dengel. Histogram-based outlier score (HBOS): A fast unsupervised anomaly detection algorithm.KI-2012: poster and demo track, 1:59–63,

    2012

  3. [7]

    DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning.arXiv preprint arXiv:2501.12948,

    Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning.arXiv preprint arXiv:2501.12948,

    Pith/arXiv arXiv

  4. [11]

    Jun Liu, Chaoyun Zhang, Jiaxu Qian, Minghua Ma, Si Qin, Chetan Bansal, Qingwei Lin, Saravan Rajmohan, and Dongmei Zhang

    10 Preprint. Jun Liu, Chaoyun Zhang, Jiaxu Qian, Minghua Ma, Si Qin, Chetan Bansal, Qingwei Lin, Saravan Rajmohan, and Dongmei Zhang. Large language models can deliver accurate and interpretable time series anomaly detection. InProceedings of the 31st ACM SIGKDD Conference on Knowl- edge Discovery and Data Mining V . 2, pp. 4623–4634, 2025a. Zijia Liu, Pe...

    Pith/arXiv arXiv

  5. [12]

    TAB: Unified benchmarking of time series anomaly detection methods.arXiv preprint arXiv:2506.18046,

    Xiangfei Qiu, Zhe Li, Wanghui Qiu, Shiyan Hu, Lekui Zhou, Xingjian Wu, Zhengyu Li, Chenjuan Guo, Aoying Zhou, Zhenli Sheng, et al. TAB: Unified benchmarking of time series anomaly detection methods.arXiv preprint arXiv:2506.18046,

    Pith/arXiv arXiv

  6. [14]

    Towards a general time series anomaly detector with adaptive bottlenecks and dual adversarial decoders.arXiv preprint arXiv:2405.15273,

    Qichao Shentu, Beibu Li, Kai Zhao, Yang Shu, Zhongwen Rao, Lujia Pan, Bin Yang, and Chen- juan Guo. Towards a general time series anomaly detector with adaptive bottlenecks and dual adversarial decoders.arXiv preprint arXiv:2405.15273,

    Pith/arXiv arXiv

  7. [15]

    Inferring events from time series using language models.arXiv preprint arXiv:2503.14190,

    Mingtian Tan, Mike A Merrill, Zack Gottesman, Tim Althoff, David Evans, and Tom Hartvigsen. Inferring events from time series using language models.arXiv preprint arXiv:2503.14190,

    Pith/arXiv arXiv

  8. [16]

    SWE-RL: Advancing LLM reasoning via reinforcement learning on open software evolution.arXiv preprint arXiv:2502.18449,

    Yuxiang Wei, Olivier Duchenne, Jade Copet, Quentin Carbonneaux, Lingming Zhang, Daniel Fried, Gabriel Synnaeve, Rishabh Singh, and Sida I Wang. SWE-RL: Advancing LLM reasoning via reinforcement learning on open software evolution.arXiv preprint arXiv:2502.18449,

    Pith/arXiv arXiv

  9. [17]

    ChatTS: Aligning time series with LLMs via synthetic data for enhanced understanding and reasoning.arXiv preprint arXiv:2412.03104,

    Zhe Xie, Zeyan Li, Xiao He, Longlong Xu, Xidao Wen, Tieying Zhang, Jianjun Chen, Rui Shi, and Dan Pei. ChatTS: Aligning time series with LLMs via synthetic data for enhanced understanding and reasoning.arXiv preprint arXiv:2412.03104,

    arXiv

  10. [18]

    Jiehui Xu, Haixu Wu, Jianmin Wang, and Mingsheng Long

    11 Preprint. Jiehui Xu, Haixu Wu, Jianmin Wang, and Mingsheng Long. Anomaly transformer: Time series anomaly detection with association discrepancy.arXiv preprint arXiv:2110.02642,

    Pith/arXiv arXiv

  11. [19]

    Can multimodal LLMs perform time series anomaly detection?arXiv preprint arXiv:2502.17812,

    Xiongxiao Xu, Haoran Wang, Yueqing Liang, Philip S Yu, Yue Zhao, and Kai Shu. Can multimodal LLMs perform time series anomaly detection?arXiv preprint arXiv:2502.17812,

    arXiv

  12. [20]

    Time-RA: Towards time series reasoning for anomaly with LLM feedback.arXiv preprint arXiv:2507.15066,

    Yiyuan Yang, Zichuan Liu, Lei Song, Kai Ying, Zhiguang Wang, Tom Bamford, Svitlana Vyetrenko, Jiang Bian, and Qingsong Wen. Time-RA: Towards time series reasoning for anomaly with LLM feedback.arXiv preprint arXiv:2507.15066,

    Pith/arXiv arXiv

  13. [21]

    Matrix profile I: all pairs similarity joins for time series: a unifying view that includes motifs, discords and shapelets

    Chin-Chia Michael Yeh, Yan Zhu, Liudmila Ulanova, Nurjahan Begum, Yifei Ding, Hoang Anh Dau, Diego Furtado Silva, Abdullah Mueen, and Eamonn Keogh. Matrix profile I: all pairs similarity joins for time series: a unifying view that includes motifs, discords and shapelets. In 2016 IEEE 16th international conference on data mining (ICDM), pp. 1317–1322. Ieee,

    2016

  14. [24]

    See it, think it, sorted: Large multimodal models are few-shot time series anomaly analyzers.arXiv preprint arXiv:2411.02465,

    Jiaxin Zhuang, Leon Yan, Zhenwei Zhang, Ruiqi Wang, Jiawei Zhang, and Yuantao Gu. See it, think it, sorted: Large multimodal models are few-shot time series anomaly analyzers.arXiv preprint arXiv:2411.02465,

    Pith/arXiv arXiv

  15. [25]

    3https://github.com/langfengQ/TimeMaster 17 Preprint

    to ensure consistency and fairness. 3https://github.com/langfengQ/TimeMaster 17 Preprint. intervals segment point amplitude timestamp value spike around (d) TimerPO Top Words expected value compared typical spikedeviation (d) GRPO Top Words(a) GRPO Alignment (b) TimerPO Alignment significant sudden ExpCoT AnomSeer Figure 12: Comparison of distributional a...

    2025

  16. [1999]

    DeepSeekMath: Pushing the limits of mathematical reasoning in open language models.arXiv preprint arXiv:2402.03300,

    Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Y Wu, et al. DeepSeekMath: Pushing the limits of mathematical reasoning in open language models.arXiv preprint arXiv:2402.03300,

    Pith/arXiv arXiv

  17. [2008]

    A picture is worth a thousand numbers: Enabling LLMs reason about time series via visualization.arXiv preprint arXiv:2411.06018,

    Haoxin Liu, Chenghao Liu, and B Aditya Prakash. A picture is worth a thousand numbers: Enabling LLMs reason about time series via visualization.arXiv preprint arXiv:2411.06018,

    Pith/arXiv arXiv

  18. [2010]

    Synergizing large language models and task-specific models for time series anomaly detection.arXiv preprint arXiv:2501.05675,

    Feiyi Chen, Leilei Zhang, Guansong Pang, Roger Zimmermann, and Shuiguang Deng. Synergizing large language models and task-specific models for time series anomaly detection.arXiv preprint arXiv:2501.05675,

    arXiv

  19. [2012]

    MOMENT: A family of open time-series foundation models.arXiv preprint arXiv:2402.03885,

    Mononito Goswami, Konrad Szafer, Arjun Choudhry, Yifu Cai, Shuo Li, and Artur Dubrawski. MOMENT: A family of open time-series foundation models.arXiv preprint arXiv:2402.03885,

    Pith/arXiv arXiv

  20. [2016]

    TimesBERT: A BERT-style foundation model for time series understanding.arXiv preprint arXiv:2502.21245, 2025a

    Haoran Zhang, Yong Liu, Yunzhong Qiu, Haixuan Liu, Zhongyi Pei, Jianmin Wang, and Mingsheng Long. TimesBERT: A BERT-style foundation model for time series understanding.arXiv preprint arXiv:2502.21245, 2025a. Junru Zhang, Lang Feng, Xu Guo, Yuhan Wu, Yabo Dong, and Duanqing Xu. TimeMaster: Training time-series multimodal LLMs to reason via reinforcement l...

    Pith/arXiv arXiv

  21. [2018]

    Position: Empowering time series reasoning with multi- modal LLMs.arXiv preprint arXiv:2502.01477,

    Yaxuan Kong, Yiyuan Yang, Shiyu Wang, Chenghao Liu, Yuxuan Liang, Ming Jin, Stefan Zohren, Dan Pei, Yan Liu, and Qingsong Wen. Position: Empowering time series reasoning with multi- modal LLMs.arXiv preprint arXiv:2502.01477,

    arXiv

  22. [2021]

    Displacement interpolation using lagrangian mass transport

    Nicolas Bonneel, Michiel Van De Panne, Sylvain Paris, and Wolfgang Heidrich. Displacement interpolation using lagrangian mass transport. InProceedings of the 2011 SIGGRAPH Asia con- ference, pp. 1–12,

    2011

  23. [2023]

    Can LLMs understand time series anomalies?arXiv preprint arXiv:2410.05440,

    Zihao Zhou and Rose Yu. Can LLMs understand time series anomalies?arXiv preprint arXiv:2410.05440,

    Pith/arXiv arXiv

  24. [2024]

    Qwen2.5-VL technical report.arXiv preprint arXiv:2502.13923,

    Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, et al. Qwen2.5-VL technical report.arXiv preprint arXiv:2502.13923,

    Pith/arXiv arXiv

  25. [2025]

    Harnessing vision-language models for time series anomaly detection.arXiv preprint arXiv:2506.06836,

    Zelin He, Sarah Alnegheimish, and Matthew Reimherr. Harnessing vision-language models for time series anomaly detection.arXiv preprint arXiv:2506.06836,

    arXiv