pith. sign in

arxiv: 2509.12089 · v6 · submitted 2025-09-15 · 📡 eess.SP · cs.CL

RadarPLM: Adapting Pre-trained Language Models for Marine Radar Target Detection by Selective Fine-tuning

Qiying Hu , Yaowen Li , Shengyi Zhang , Chuan Huang , Yu Liu , You He This is my paper

Pith reviewed 2026-05-18 15:55 UTC · model grok-4.3

classification 📡 eess.SP cs.CL
keywords pre-trained language modelsmarine radartarget detectionselective fine-tuninglow SCRadaptation modulesignal processingoverfitting
0
0 comments X p. Extension

The pith

Adapting pre-trained language models with selective fine-tuning enables effective marine radar target detection even in low signal-to-clutter conditions.

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

This paper shows how pre-trained language models originally developed for text can be repurposed for detecting targets in marine radar signals. Direct fine-tuning of these models on radar data is computationally expensive and tends to overfit, especially when the signal-to-clutter ratio is low. The authors introduce a lightweight adaptation module that adds radar-specific adjustments while keeping the model's broad pre-trained knowledge intact. They pair it with a selective fine-tuning strategy that scores feature patches by their online learning value and focuses updates on the most useful ones, steering away from noise or trivial patterns. A final autoencoder-based classification head is retrained to sharpen the output. Real-world tests show at least a 6.35 percent gain in average detection performance under challenging low-SCR conditions and strong results even with limited training samples.

Core claim

The RadarPLM framework adapts pre-trained language models for marine radar target detection by inserting a lightweight adaptation module for efficient fine-tuning, applying a selective fine-tuning strategy that optimizes feature patches according to their online-evaluated learning values to emphasize generalizable patterns, and retraining a binary classification head based on an autoencoder network. This combination preserves the universal knowledge captured in the language model while reducing both computational cost and overfitting to noisy or simple patterns that commonly appear in low-SCR radar environments, resulting in higher detection accuracy on real-world datasets.

What carries the argument

The selective fine-tuning strategy that optimizes different feature patches based on their online-evaluated learning values to guide the model toward generalizable patterns.

If this is right

  • The framework achieves a minimum 6.35% gain in average detection performance under low SCR conditions when using sequence features.
  • It delivers highly significant average performance gains over prior methods under small-sample training conditions.
  • The lightweight adaptation module enables computationally efficient fine-tuning while preserving the pre-trained model's general knowledge.
  • The selective strategy reduces model overfitting to noisy, anomalous, or overly simple patterns during optimization.
  • Integration of pre-trained language models proves effective for radar signal processing tasks.

Where Pith is reading between the lines

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

  • The same lightweight-plus-selective approach could transfer to other noisy time-series detection problems such as sonar or vibration monitoring.
  • Evaluating learning values online during fine-tuning may serve as a general technique to improve robustness when adapting large models to limited or noisy data in any domain.
  • The success with sequence features suggests that language-model-style sequential processing captures useful structure in radar returns beyond what hand-crafted signal features provide.

Load-bearing premise

The online-evaluated learning values of feature patches reliably distinguish generalizable patterns from noise or overly simple artifacts in low-SCR radar data.

What would settle it

Removing the selective component or replacing it with random patch selection and finding that detection performance no longer improves or falls below baseline methods on the same real-world low-SCR radar datasets.

Figures

Figures reproduced from arXiv: 2509.12089 by Chuan Huang, Qiying Hu, Shengyi Zhang, Yaowen Li, You He, Yu Liu.

Figure 1
Figure 1. Figure 1: Overview of our RadarPLM framework. Patch the sequence features Linear Projection+Position Embedding 1 Transformer Encoder 2 3 4 5 6 Linear Layer*2 [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the reference model. IP feature DSE feature SMS feature Amp feature DP feature Patching [5, N] [K, L] 5 ... K [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: An illustration of the patching operation. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: IP, DSE, SMS, Amp, and DP features for target and sea clutter echo signal on IPIX database. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Illustration of the LoRA fine-tuning process. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Comparison of model detection performance on various datasets. [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of detection rates between the proposed preference-aware loss and the conventional cross-entropy loss. We compare detection rates across [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
read the original abstract

Recent advances in pre-trained language models (PLMs) have demonstrated their capabilities in capturing universal knowledge, making them promising for radar signal processing applications. Nevertheless, directly fine-tuning PLMs on radar signals is both computationally expensive and prone to overfitting, particularly in low signal-to-clutter ratio (SCR) environments. To mitigate both issues, an effective fine-tuning framework for PLM-based marine radar target detection is proposed. First, we design a lightweight adaptation module, enabling computationally efficient fine-tuning while preserving the pre-trained model's general knowledge. Second, an effective selective fine-tuning strategy is developed to selectively optimize different feature patches based on their online-evaluated learning values, guiding the model to concentrate on those generalizable feature patterns and significantly reducing model overfitting to nosiy, anomalous, or overly simple patterns during optimization. Finally, a binary classification head is retrained based on autoencoder network to further enhance detection performance. Evaluations on real-world radar datasets highlight that the proposed RadarPLM framework considerably outperforms existing models, achieving a minimum of 6.35% gain in average detection performance under challenging low SCR conditions when using sequence features. In particular, under small-sample training conditions, RadarPLM also achieves highly significant average performance gains over prior methods, demonstrating the effectiveness of integrating the PLM.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes RadarPLM, a framework adapting pre-trained language models for marine radar target detection. It introduces a lightweight adaptation module for efficient fine-tuning, a selective fine-tuning strategy that assigns online-evaluated learning values to feature patches and optimizes only high-value ones to reduce overfitting in low-SCR conditions, and a retrained binary classification head based on an autoencoder. Evaluations on real-world radar datasets report a minimum 6.35% gain in average detection performance under low SCR with sequence features, plus gains in small-sample regimes.

Significance. If the central empirical claims hold after addressing the noted gaps, the work would usefully demonstrate transfer of PLM capabilities to radar signal processing, particularly for low-SCR marine target detection where data are noisy and samples limited. The combination of lightweight adaptation with selectivity offers a practical route to domain adaptation without full retraining, and the reported gains on real datasets plus small-sample results constitute concrete evidence of applicability. These elements could inform future cross-domain PLM uses in sensing applications.

major comments (2)
  1. [§4] §4 (Experimental Evaluation): The headline claim of a minimum 6.35% gain under low SCR is attributed to the selective fine-tuning strategy, yet no ablation is presented that removes selectivity while retaining the identical lightweight adaptation module and autoencoder head. Without this controlled comparison, the contribution of patch selection versus the adapter or head cannot be isolated, weakening the causal link to reduced overfitting.
  2. [§3.2] §3.2 (Selective Fine-Tuning Strategy): The mechanism for computing 'online-evaluated learning values' for feature patches and the precise selection threshold or optimization rule are described only at a high level. This absence of equations, pseudocode, or diagnostic statistics (e.g., distribution of selected vs. discarded patches or loss curves) makes it impossible to verify that low-value patches correspond to noise rather than merely acting as an additional regularizer.
minor comments (2)
  1. [Abstract] Abstract: 'nosiy' is a typographical error and should read 'noisy'.
  2. The manuscript would benefit from a dedicated limitations or failure-mode subsection discussing conditions under which the selective strategy might still overfit or degrade performance.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments on our manuscript. We have carefully reviewed the feedback and provide point-by-point responses below. Where appropriate, we commit to revisions that will strengthen the presentation of our contributions without altering the core claims.

read point-by-point responses

  1. Referee: [§4] §4 (Experimental Evaluation): The headline claim of a minimum 6.35% gain under low SCR is attributed to the selective fine-tuning strategy, yet no ablation is presented that removes selectivity while retaining the identical lightweight adaptation module and autoencoder head. Without this controlled comparison, the contribution of patch selection versus the adapter or head cannot be isolated, weakening the causal link to reduced overfitting.

    Authors: We acknowledge that an explicit ablation study isolating the selective fine-tuning component—while holding the lightweight adaptation module and autoencoder head fixed—would provide stronger evidence for its specific role in mitigating overfitting. The current evaluations demonstrate overall gains of the full RadarPLM framework relative to prior methods on real-world datasets, including under low-SCR and small-sample regimes. To directly address this point, we will add the requested controlled ablation to Section 4 in the revised manuscript. revision: yes

  2. Referee: [§3.2] §3.2 (Selective Fine-Tuning Strategy): The mechanism for computing 'online-evaluated learning values' for feature patches and the precise selection threshold or optimization rule are described only at a high level. This absence of equations, pseudocode, or diagnostic statistics (e.g., distribution of selected vs. discarded patches or loss curves) makes it impossible to verify that low-value patches correspond to noise rather than merely acting as an additional regularizer.

    Authors: We agree that the description in Section 3.2 would benefit from greater technical detail to enable verification. In the revised manuscript, we will expand this section to include the exact equations governing the computation of online-evaluated learning values, the selection threshold and optimization rule, pseudocode for the procedure, and diagnostic statistics such as distributions of selected versus discarded patches and comparative loss curves. These additions will clarify the mechanism's operation. revision: yes

Circularity Check

0 steps flagged

No circularity in empirical adaptation framework

full rationale

The paper describes a practical fine-tuning pipeline for PLMs on radar data, consisting of a lightweight adapter, a selective optimization step that uses online learning-value scores on feature patches, and a retrained autoencoder-based head. All reported gains (including the 6.35 % low-SCR improvement) are obtained from direct experimental comparison on real-world datasets rather than from any closed-form derivation, fitted parameter, or self-referential equation that would make the outcome identical to the input by construction. No mathematical chain, uniqueness theorem, or ansatz is invoked that reduces the central performance claim to a tautology; the work therefore remains self-contained empirical validation.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The abstract provides limited technical detail, so the ledger is necessarily incomplete. The framework assumes PLMs encode transferable knowledge and that learning-value scoring can be computed reliably online.

axioms (1)
  • domain assumption Pre-trained language models capture universal knowledge that can transfer to radar signal processing.
    Stated in the opening sentence of the abstract as the motivation for applying PLMs to radar.

pith-pipeline@v0.9.0 · 5774 in / 1236 out tokens · 33594 ms · 2026-05-18T15:55:52.056698+00:00 · methodology

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

42 extracted references · 42 canonical work pages · 4 internal anchors

  1. [1]

    Detection and extraction of target with micromotion in spiky sea clutter via short-time fractional fourier transform,

    X. Chen, J. Guan, Z. Bao, and Y . He, “Detection and extraction of target with micromotion in spiky sea clutter via short-time fractional fourier transform,”IEEE Transactions on Geoscience and Remote Sensing, vol. 52, no. 2, pp. 1002–1018, 2013

    work page 2013

  2. [2]

    Tri-feature-based detection of floating small targets in sea clutter,

    P.-L. Shui, D.-C. Li, and S.-W. Xu, “Tri-feature-based detection of floating small targets in sea clutter,”IEEE Transactions on Aerospace and Electronic Systems, vol. 50, no. 2, pp. 1416–1430, 2014

    work page 2014

  3. [3]

    Sea-surface floating small target detection by one-class classifier in time-frequency feature space,

    S.-N. Shi and P.-L. Shui, “Sea-surface floating small target detection by one-class classifier in time-frequency feature space,”IEEE Transactions on Geoscience and Remote Sensing, vol. 56, no. 11, pp. 6395–6411, 2018

    work page 2018

  4. [4]

    Floating small target detection in sea clutter based on multifeature angle variance,

    X. Bai, S. Xu, J. Zhu, Z. Guo, and P. Shui, “Floating small target detection in sea clutter based on multifeature angle variance,”IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 16, pp. 9422–9436, 2023

    work page 2023

  5. [5]

    Eigenvalues-based detector design for radar small floating target detection in sea clutter,

    W. Zhao, M. Jin, G. Cui, and Y . Wang, “Eigenvalues-based detector design for radar small floating target detection in sea clutter,”IEEE Geoscience and Remote Sensing Letters, vol. 19, pp. 1–5, 2021

    work page 2021

  6. [6]

    Sea-surface floating small target detection by multifeature detector based on isolation forest,

    S. Xu, J. Zhu, J. Jiang, and P. Shui, “Sea-surface floating small target detection by multifeature detector based on isolation forest,”IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 14, pp. 704–715, 2020

    work page 2020

  7. [7]

    False-alarm-controllable radar detection for marine target based on multi features fusion via cnns,

    X. Chen, N. Su, Y . Huang, and J. Guan, “False-alarm-controllable radar detection for marine target based on multi features fusion via cnns,” IEEE Sensors Journal, vol. 21, no. 7, pp. 9099–9111, 2021

    work page 2021

  8. [8]

    A false alarm controllable detection method based on cnn for sea-surface small targets,

    Q. Qu, Y .-L. Wang, W. Liu, and B. Li, “A false alarm controllable detection method based on cnn for sea-surface small targets,”IEEE Geoscience and Remote Sensing Letters, vol. 19, pp. 1–5, 2022

    work page 2022

  9. [9]

    Marine radar small target classification based on block-whitened time–frequency spectrogram and pre-trained cnn,

    S. Xu, H. Ru, D. Li, P. Shui, and J. Xue, “Marine radar small target classification based on block-whitened time–frequency spectrogram and pre-trained cnn,”IEEE Transactions on Geoscience and Remote Sensing, vol. 61, pp. 1–11, 2023

    work page 2023

  10. [10]

    Target detection in sea clutter via contrastive learning,

    S. Xia, Y . Kong, K. Xiong, and G. Cui, “Target detection in sea clutter via contrastive learning,”IEEE Transactions on Instrumentation and Measurement, vol. 72, pp. 1–13, 2023

    work page 2023

  11. [11]

    Nonhomoge- neous sea clutter suppression using complex-valued u-net model,

    Y . Wang, W. Zhao, X. Wang, J. Chen, H. Li, and G. Cui, “Nonhomoge- neous sea clutter suppression using complex-valued u-net model,”IEEE Geoscience and Remote Sensing Letters, vol. 19, pp. 1–5, 2022

    work page 2022

  12. [12]

    Maritime radar target detection in sea clutter based on cnn with dual-perspective attention,

    J. Wang and S. Li, “Maritime radar target detection in sea clutter based on cnn with dual-perspective attention,”IEEE Geoscience and Remote Sensing Letters, vol. 20, pp. 1–5, 2022

    work page 2022

  13. [14]

    Sequence-feature detection of small targets in sea clutter based on bi-lstm,

    H. Wan, X. Tian, J. Liang, and X. Shen, “Sequence-feature detection of small targets in sea clutter based on bi-lstm,”IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1–11, 2022

    work page 2022

  14. [15]

    Maritime target detection based on radar graph data and graph convolutional network,

    N. Su, X. Chen, J. Guan, and Y . Huang, “Maritime target detection based on radar graph data and graph convolutional network,”IEEE Geoscience and Remote Sensing Letters, vol. 19, pp. 1–5, 2021

    work page 2021

  15. [16]

    Marine small floating target detection method based on fusion weight and graph dynamic attention mechanism,

    H. Ru, S. Xu, Q. He, and P. Shui, “Marine small floating target detection method based on fusion weight and graph dynamic attention mechanism,”IEEE Transactions on Geoscience and Remote Sensing, vol. 61, pp. 1–11, 2023

    work page 2023

  16. [17]

    Radar maritime target detection via spatial–temporal feature attention graph convolutional network,

    N. Su, X. Chen, J. Guan, Y . Huang, X. Wang, and Y . Xue, “Radar maritime target detection via spatial–temporal feature attention graph convolutional network,”IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1–15, 2024

    work page 2024

  17. [18]

    One fits all: Power general time series analysis by pretrained lm,

    T. Zhou, P. Niu, L. Sun, R. Jinet al., “One fits all: Power general time series analysis by pretrained lm,”Advances in neural information processing systems, vol. 36, pp. 43 322–43 355, 2023

    work page 2023

  18. [19]

    Unitime: A language-empowered unified model for cross-domain time series forecasting,

    X. Liu, J. Hu, Y . Li, S. Diao, Y . Liang, B. Hooi, and R. Zimmermann, “Unitime: A language-empowered unified model for cross-domain time series forecasting,” inProceedings of the ACM on Web Conference 2024, 2024, pp. 4095–4106

    work page 2024

  19. [20]

    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. Panet al., “Time-llm: Time series forecasting by re- programming large language models,”arXiv preprint arXiv:2310.01728, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  20. [21]

    Large language model enabled multi-task physical layer network,

    T. Zheng and L. Dai, “Large language model enabled multi-task physical layer network,”arXiv preprint arXiv:2412.20772, 2024

    work page arXiv 2024

  21. [22]

    Llm-empowered near-field communica- tions for low-altitude economy,

    Z. Xu, T. Zheng, and L. Dai, “Llm-empowered near-field communica- tions for low-altitude economy,”IEEE Transactions on Communications, pp. 1–1, 2025

    work page 2025

  22. [23]

    Bridging the modality gap: Enhancing channel prediction with semantically aligned llms and knowledge distillation,

    Z. Li, Q. Yang, Z. Xiong, Z. Shi, and T. Q. Quek, “Bridging the modality gap: Enhancing channel prediction with semantically aligned llms and knowledge distillation,”arXiv preprint arXiv:2505.12729, 2025

    work page arXiv 2025

  23. [24]

    Llm4cp: Adapting large language models for channel prediction,

    B. Liu, X. Liu, S. Gao, X. Cheng, and L. Yang, “Llm4cp: Adapting large language models for channel prediction,”Journal of Communications and Information Networks, vol. 9, no. 2, pp. 113–125, 2024

    work page 2024

  24. [25]

    Rho-1: Not all tokens are what you need

    Z. Lin, Z. Gou, Y . Gong, X. Liu, Y . Shen, R. Xu, C. Lin, Y . Yang, J. Jiao, N. Duanet al., “Rho-1: Not all tokens are what you need,” arXiv preprint arXiv:2404.07965, 2024

    work page arXiv 2024

  25. [26]

    Prioritized training on points that are learnable, worth learning, and not yet learnt,

    S. Mindermann, J. M. Brauner, M. T. Razzak, M. Sharma, A. Kirsch, W. Xu, B. H ¨oltgen, A. N. Gomez, A. Morisot, S. Farquharet al., “Prioritized training on points that are learnable, worth learning, and not yet learnt,” inInternational Conference on Machine Learning. PMLR, 2022, pp. 15 630–15 649

    work page 2022

  26. [27]

    When marine radar target detection meets pretrained large language models,

    Q. Hu, L. Zhang, X. Wang, G. Li, Y . Liu, and X.-P. Zhang, “When marine radar target detection meets pretrained large language models,” in IGARSS 2025-2025 IEEE International Geoscience and Remote Sensing Symposium, 2025, in press

    work page 2025

  27. [28]

    Maritime radar target detection model self-evolution based on semisupervised learning,

    J. Wang and S. Li, “Maritime radar target detection model self-evolution based on semisupervised learning,”IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1–11, 2023

    work page 2023

  28. [29]

    Radar detection self-evolution based on incremental learning in time-varying clutter environments,

    Y . Wang, X. Chen, X. Wang, C. Zang, W. Zhao, and G. Cui, “Radar detection self-evolution based on incremental learning in time-varying clutter environments,”IEEE Transactions on Instrumentation and Mea- surement, vol. 73, pp. 1–13, 2024

    work page 2024

  29. [30]

    A survey on evaluation of large language models,

    Y . Chang, X. Wang, J. Wang, Y . Wu, L. Yang, K. Zhu, H. Chen, X. Yi, C. Wang, Y . Wanget al., “A survey on evaluation of large language models,”ACM transactions on intelligent systems and technology, vol. 15, no. 3, pp. 1–45, 2024. IEEE JOURNAL 14

    work page 2024

  30. [31]

    Language models are unsupervised multitask learners,

    A. Radford, J. Wu, R. Child, D. Luan, D. Amodei, I. Sutskeveret al., “Language models are unsupervised multitask learners,”OpenAI blog, vol. 1, no. 8, p. 9, 2019

    work page 2019

  31. [32]

    Beam predic- tion based on large language models,

    Y . Sheng, K. Huang, L. Liang, P. Liu, S. Jin, and G. Y . Li, “Beam predic- tion based on large language models,”arXiv preprint arXiv:2408.08707, 2024

    work page arXiv 2024

  32. [33]

    Iot-llm: Enhancing real- world iot task reasoning with large language models,

    T. An, Y . Zhou, H. Zou, and J. Yang, “Iot-llm: Enhancing real- world iot task reasoning with large language models,”arXiv preprint arXiv:2410.02429, 2024

    work page arXiv 2024

  33. [34]

    Penetrative ai: Making llms comprehend the physical world,

    H. Xu, L. Han, Q. Yang, M. Li, and M. Srivastava, “Penetrative ai: Making llms comprehend the physical world,” inProceedings of the 25th International Workshop on Mobile Computing Systems and Applications, 2024, pp. 1–7

    work page 2024

  34. [35]

    Are language models actually useful for time series forecasting?

    M. Tan, M. Merrill, V . Gupta, T. Althoff, and T. Hartvigsen, “Are language models actually useful for time series forecasting?”Advances in Neural Information Processing Systems, vol. 37, pp. 60 162–60 191, 2024

    work page 2024

  35. [36]

    Attention Is All You Need

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,”arXiv preprint arXiv:1706.03762, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  36. [37]

    Lora: Low-rank adaptation of large language models

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, W. Chenet al., “Lora: Low-rank adaptation of large language models.” ICLR, vol. 1, no. 2, p. 3, 2022

    work page 2022

  37. [38]

    Small target classification of maritime radars using standardized time-frequency distribution images and residual networks with improved focal loss,

    J.-Y . Li, X.-J. Zhang, Z.-J. Zhao, P.-L. Shui, and S.-W. Xu, “Small target classification of maritime radars using standardized time-frequency distribution images and residual networks with improved focal loss,” IEEE Transactions on Instrumentation and Measurement, vol. 74, pp. 1–17, 2025

    work page 2025

  38. [39]

    Neural network based false alarm controllable detector for weak targets via a differentiable np loss,

    Q. Qu, H. Chen, W. Liu, B. Li, and Y .-L. Wang, “Neural network based false alarm controllable detector for weak targets via a differentiable np loss,”IEEE Transactions on Aerospace and Electronic Systems, pp. 1–13, 2025

    work page 2025

  39. [40]

    Multi-task learning using uncer- tainty to weigh losses for scene geometry and semantics,

    A. Kendall, Y . Gal, and R. Cipolla, “Multi-task learning using uncer- tainty to weigh losses for scene geometry and semantics,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 7482–7491

    work page 2018

  40. [41]

    Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation

    K. Cho, “Learning phrase representations using rnn encoder-decoder for statistical machine translation,”arXiv preprint arXiv:1406.1078, 2014

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  41. [42]

    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,”arXiv preprint arXiv:2211.14730, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  42. [43]

    Enhanced cnn-based small target detection in sea clutter with controllable false alarm,

    Q. Qu, W. Liu, J. Wang, B. Li, N. Liu, and Y .-L. Wang, “Enhanced cnn-based small target detection in sea clutter with controllable false alarm,”IEEE Sensors Journal, vol. 23, no. 9, pp. 10 193–10 205, 2023

    work page 2023