arxiv: 2605.03383 · v1 · submitted 2026-05-05 · 💻 cs.AI

Recognition: unknown

GeoDecider: A Coarse-to-Fine Agentic Workflow for Explainable Lithology Classification

Jiahao Wang , Mingyue Cheng , Yitong Zhou , Qingyang Mao , Xiaoyu Tao , Qi Liu , Enhong Chen

Authors on Pith no claims yet

Pith reviewed 2026-05-07 16:40 UTC · model grok-4.3

classification 💻 cs.AI

keywords lithology classificationwell-logginglarge language modelsagentic workflowgeological refinementcoarse-to-fine reasoningexplainable predictionssubsurface analysis

0 comments

The pith

GeoDecider reformulates lithology classification as a multi-stage LLM workflow that adds geological consistency and explainability to well-log analysis.

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

The paper aims to replace single-pass classification models with a structured process that mirrors how geologists use principles, knowledge, and tools. GeoDecider breaks the task into coarse classification from a base model, tool-augmented refinement using context and neighbors, and final geological post-processing. This training-free approach with large language models seeks both higher accuracy and interpretable outputs. A reader would care because lithology identification feeds directly into reservoir characterization and other subsurface decisions. If the claim holds, the method would deliver predictions that align better with real geological sequences while using less computation than end-to-end alternatives.

Core claim

GeoDecider organizes lithology classification into a coarse-to-fine agentic workflow with three stages: base classifier-guided coarse classification that supplies an initial reference, tool-augmented reasoning that applies contextual analysis and neighbor retrieval for precision, and geological refinement that enforces consistency on the final output. Large language models execute the workflow without additional training. On four benchmarks the approach outperforms representative baselines while producing geologically interpretable predictions and a favorable accuracy-efficiency trade-off.

What carries the argument

The coarse-to-fine agentic workflow that sequences initial rough classification, tool-driven refinement, and geological consistency enforcement to progressively improve results from large language models.

Load-bearing premise

Large language models can reliably apply geological principles and tool outputs to refine classifications without introducing new errors or inconsistencies.

What would settle it

A dataset of well logs containing known geological sequences where the workflow either produces impossible rock-type transitions or fails to match baseline accuracy.

Figures

Figures reproduced from arXiv: 2605.03383 by Enhong Chen, Jiahao Wang, Mingyue Cheng, Qi Liu, Qingyang Mao, Xiaoyu Tao, Yitong Zhou.

**Figure 1.** Figure 1: Example of multivariate well logs from various view at source ↗

**Figure 2.** Figure 2: The illustration of GeoDecider, a coarse-to-fine framework for lithology classification. view at source ↗

**Figure 3.** Figure 3: Performance comparison between lightweight base view at source ↗

**Figure 4.** Figure 4: The improvement of F1 and the number of process view at source ↗

**Figure 6.** Figure 6: A case of how GeoDecider performs lithology classification via pre-classified and tool-augmented reasoning. view at source ↗

**Figure 7.** Figure 7: Sensitivity analysis of GeoDecider to the tempera view at source ↗

read the original abstract

Lithology classification aims to infer subsurface rock types from well-logging signals, supporting downstream applications like reservoir characterization. Despite substantial progress, most existing methods still treat lithology classification as a single-pass classification task. In contrast, practical experts incorporate geological principles, external knowledge, and tool-use capabilities to perform accurate classification. In this work, we propose GeoDecider, a coarse-to-fine agentic workflow that enables accurate and explainable lithology classification through training-free use of large language models (LLMs). GeoDecider reformulates lithology classification as an expert-like structured process and organizes it into a multi-stage workflow involving coarse-to-fine reasoning. Specifically, GeoDecider includes the following stages: (1) base classifier-guided coarse classification, which uses a pre-trained classifier to provide a rough reference for downstream tasks, thus reducing the overall cost of downstream reasoning, (2) tool-augmented reasoning, which utilizes several tools such as contextual analysis and neighbor retrieval to achieve finer and more precise classifications, (3) geological refinement, which post-processes the final results to enforce geological consistency. Experiments on four benchmarks show that GeoDecider outperforms representative baselines. Further analysis demonstrates that the proposed framework produces geologically interpretable predictions while achieving a better trade-off between classification performance and inference efficiency.

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.

Desk Editor's Note private letter to a colleague

GeoDecider offers a sensible staged LLM workflow for lithology classification but its outperformance claims rest on details not visible in the abstract.

read the letter

The main thing to know is that this paper describes a three-stage agentic process—base classifier for a cheap coarse pass, tool-augmented reasoning for finer calls, and geological refinement for consistency—rather than another single-pass model. That structure is new for the domain and directly addresses how experts actually work with logs, external knowledge, and rules. It is training-free, which keeps the cost low and makes the outputs more traceable than black-box classifiers. The paper does a clean job laying out why each stage exists and how they chain together to improve both accuracy and interpretability. The efficiency angle is also sensible: the coarse step prunes the LLM calls that follow. Those are real contributions on the method side. The soft spots are straightforward. The abstract states that the workflow beats baselines on four benchmarks and shows a better performance-efficiency trade-off, yet it supplies no numbers, no baseline descriptions, no statistical tests, and no breakdown of where the gains come from or where errors remain. The refinement stage is the load-bearing piece, and nothing in the visible text shows how often it overrides the coarse prediction, whether those overrides are correct, or whether it introduces new inconsistencies. The stress-test concern about the LLM refinement step potentially adding errors without validation is therefore still live. This paper is aimed at people working on applied geoscience AI or agentic systems for domain tasks. A reader looking for concrete workflow ideas will find usable structure here even if the numbers are not yet convincing. It deserves a serious referee because the proposal is coherent and the gap it targets is real; the experiments simply need to be shown in full with proper controls and analysis before the claims can be assessed.

Referee Report

2 major / 1 minor

Summary. The manuscript proposes GeoDecider, a training-free coarse-to-fine agentic workflow for lithology classification from well-logging signals using large language models. The method organizes the task into three stages: (1) base classifier-guided coarse classification to provide an initial low-cost reference, (2) tool-augmented reasoning employing contextual analysis and neighbor retrieval for refinement, and (3) geological refinement to post-process outputs for consistency with geological principles. The central claims are that this workflow outperforms representative baselines on four benchmarks, yields geologically interpretable predictions, and achieves a superior performance-efficiency trade-off.

Significance. If the empirical results and refinement validation hold, the work could meaningfully advance agentic LLM applications in scientific domains by showing how structured multi-stage reasoning can incorporate external knowledge and domain principles for improved explainability and accuracy. The training-free design and explicit separation of coarse and fine stages are potential strengths for efficiency. However, the absence of supporting metrics and validation details in the current description makes it difficult to gauge the actual contribution or broader impact on geoscience AI methods.

major comments (2)

[Abstract] The abstract states that 'Experiments on four benchmarks show that GeoDecider outperforms representative baselines' and 'achieving a better trade-off between classification performance and inference efficiency,' yet supplies no quantitative metrics, baseline names or implementations, statistical significance tests, or error analysis. This omission is load-bearing for the primary empirical claim and prevents assessment of whether the reported gains are real or meaningful.
[Geological refinement stage] The geological refinement stage is presented as the final post-processing step that enforces geological consistency using principles and external knowledge, but the manuscript provides no validation of its effects, such as counts of predictions altered by refinement, expert review of those changes, or analysis of whether it overrides correct coarse predictions. Without such checks, it is impossible to confirm that this load-bearing stage produces net accuracy gains rather than introducing inconsistencies or errors.

minor comments (1)

[Abstract] The abstract refers to 'representative baselines' without naming or briefly characterizing them; adding this detail would improve readability and context even at the summary level.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which help clarify how to strengthen the presentation of our empirical claims and the validation of key components. We address each major comment below and commit to revisions that directly respond to the concerns raised.

read point-by-point responses

Referee: [Abstract] The abstract states that 'Experiments on four benchmarks show that GeoDecider outperforms representative baselines' and 'achieving a better trade-off between classification performance and inference efficiency,' yet supplies no quantitative metrics, baseline names or implementations, statistical significance tests, or error analysis. This omission is load-bearing for the primary empirical claim and prevents assessment of whether the reported gains are real or meaningful.

Authors: We agree that the abstract would benefit from greater specificity to allow readers to immediately assess the strength of the results. In the revised manuscript we will update the abstract to report concrete metrics (average accuracy/F1 improvement across the four benchmarks), name the representative baselines, and briefly reference the efficiency gains with supporting numbers drawn from the main experiments. The main text already contains the full tables, statistical significance tests, and error analysis; the abstract revision will provide a concise summary of those findings while remaining within length limits. revision: yes
Referee: [Geological refinement stage] The geological refinement stage is presented as the final post-processing step that enforces geological consistency using principles and external knowledge, but the manuscript provides no validation of its effects, such as counts of predictions altered by refinement, expert review of those changes, or analysis of whether it overrides correct coarse predictions. Without such checks, it is impossible to confirm that this load-bearing stage produces net accuracy gains rather than introducing inconsistencies or errors.

Authors: We acknowledge that explicit validation of the geological refinement stage is necessary to substantiate its contribution. We will add a dedicated analysis section (or subsection) that reports: (i) the number and percentage of predictions changed by the refinement step on each benchmark, (ii) before-and-after accuracy comparisons to quantify net gains, and (iii) representative examples illustrating the geological principles applied and whether any correct coarse predictions were altered. Where feasible we will also include qualitative expert commentary on a sample of changes; otherwise we will rely on the quantitative ablation to demonstrate that the stage improves consistency without net harm. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical workflow on external benchmarks

full rationale

The paper presents GeoDecider as a multi-stage empirical workflow (coarse classifier + tool-augmented LLM reasoning + geological refinement) for lithology classification. It is evaluated directly on four external benchmarks with reported accuracy and efficiency gains. No mathematical derivation chain, fitted parameters renamed as predictions, self-referential definitions, or load-bearing self-citations appear in the provided text. The central claims rest on experimental outperformance rather than any reduction of outputs to inputs by construction. This is the standard case of a self-contained applied method with no circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The approach rests primarily on the assumption that LLMs can perform reliable geological reasoning when given tools and context, without introducing new physical entities or many fitted parameters.

axioms (1)

domain assumption Large language models can perform contextual analysis, neighbor retrieval, and geological refinement when provided with appropriate tools and instructions.
This underpins the tool-augmented reasoning and geological refinement stages described in the abstract.

pith-pipeline@v0.9.0 · 5551 in / 1425 out tokens · 75795 ms · 2026-05-07T16:40:04.994542+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Reference graph

Works this paper leans on

72 extracted references · 12 canonical work pages · 6 internal anchors

[1]

Mohammed AM Abdullah, Ahmed A Mohammed, and Sohaib R Awad. 2024. RockDNet: Deep learning approach for lithology classification.Applied Sciences 14, 13 (2024), 5511

2024
[2]

Fatimah Alzubaidi, Peyman Mostaghimi, Pawel Swietojanski, Stuart R Clark, and Ryan T Armstrong. 2021. Automated lithology classification from drill core images using convolutional neural networks.Journal of Petroleum Science and Engineering197 (2021), 107933

2021
[3]

Roman Beloborodov, James Gunning, Marina Pervukhina, Juerg Hauser, Michael Ben Clennell, Alan Mur, and Vladimir Li. 2024. Automated lithofluid and facies classification in well logs: The rock-physics perspective.Geophysics 89, 4 (2024), MR209–MR222

2024
[4]

Filipe Beretta, AL Rodrigues, RL Peroni, and JFCL Costa. 2019. Automated lithological classification using UAV and machine learning on an open cast mine. Applied Earth Science128, 3 (2019), 79–88

2019
[5]

Thiago Santi Bressan, Marcelo Kehl de Souza, Tiago J Girelli, and Farid Chemale Junior. 2020. Evaluation of machine learning methods for lithology classification using geophysical data.Computers & Geosciences139 (2020), 104475

2020
[6]

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners.Advances in neural information processing systems33 (2020), 1877–1901

2020
[7]

Luoyuan Chen, Xingjian Wang, and Zhanbo Liu. 2025. Geological information- driven deep learning for lithology identification from well logs.Frontiers in Earth Science13 (2025), 1662760

2025
[8]

Tianqi Chen, Tong He, Michael Benesty, Vadim Khotilovich, Yuan Tang, Hyunsu Cho, Kailong Chen, Rory Mitchell, Ignacio Cano, Tianyi Zhou, et al. 2015. Xgboost: extreme gradient boosting.R package version 0.4-21, 4 (2015), 1–4

2015
[9]

Mingyue Cheng, Yiheng Chen, Qi Liu, Zhiding Liu, Yucong Luo, and Enhong Chen. 2025. InstrucTime: Advancing Time Series Classification with Multimodal Language Modeling. InProceedings of the Eighteenth ACM International Conference on Web Search and Data Mining(Hannover, Germany)(WSDM ’25). Association for Computing Machinery, New York, NY, USA, 792–800. do...

work page doi:10.1145/3701551 2025
[10]

Mingyue Cheng, Qi Liu, Zhiding Liu, Zhi Li, Yucong Luo, and Enhong Chen. 2023. FormerTime: Hierarchical Multi-Scale Representations for Multivariate Time Series Classification. InProceedings of the ACM Web Conference 2023(Austin, TX, USA)(WWW ’23). Association for Computing Machinery, New York, NY, USA, 1437–1445. doi:10.1145/3543507.3583205

work page doi:10.1145/3543507.3583205 2023
[11]

Mingyue Cheng, Qi Liu, Qingyang Mao, Yitong Zhou, Yupeng Li, Jiahao Wang, Jiaying Lin, Jiawei Cao, and Enhong Chen. 2025. A survey on table mining with large language models: Challenges, advancements and prospects. (2025)

2025
[12]

Mingyue Cheng, Zhiding Liu, Xiaoyu Tao, Qi Liu, Jintao Zhang, Tingyue Pan, Shilong Zhang, Panjing He, Xiaohan Zhang, Daoyu Wang, et al. 2025. A com- prehensive survey of time series forecasting: Concepts, challenges, and future directions.Authorea Preprints(2025)

2025
[13]

Mingyue Cheng, Jiahao Wang, Daoyu Wang, Xiaoyu Tao, Qi Liu, and Enhong Chen. 2026. Can slow-thinking llms reason over time? empirical studies in time series forecasting. InProceedings of the Nineteenth ACM International Conference on Web Search and Data Mining. 99–110

2026
[14]

Schmidt, and Geoffrey I

Angus Dempster, Daniel F. Schmidt, and Geoffrey I. Webb. 2021. MiniRocket: A Very Fast (Almost) Deterministic Transform for Time Series Classification. InProceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining(Virtual Event, Singapore)(KDD ’21). Association for Computing Machinery, New York, NY, USA, 248–257. doi:10.1145/3447548.3467231

work page doi:10.1145/3447548.3467231 2021
[15]

Shao-Qun Dong, Zhao-Hui Zhong, Xue-Hui Cui, Lian-Bo Zeng, Xu Yang, Jian-Jun Liu, Yan-Ming Sun, and Jing-Ru Hao. 2023. A deep kernel method for lithofacies identification using conventional well logs.Petroleum Science20, 3 (2023), 1411– 1428

2023
[16]

Mohamed Ali El-Omairi and Abdelkader El Garouani. 2023. A review on ad- vancements in lithological mapping utilizing machine learning algorithms and remote sensing data.Heliyon9, 9 (2023)

2023
[17]

Lutfi Eren Erdogan, Nicholas Lee, Sehoon Kim, Suhong Moon, Hiroki Furuta, Gopala Anumanchipalli, Kurt Keutzer, and Amir Gholami. 2025. Plan-and-Act: Im- proving Planning of Agents for Long-Horizon Tasks. InInternational Conference on Machine Learning. PMLR, 15419–15462

2025
[18]

Zuochun Fan, Changhao Hu, Shu Jiang, Man Li, Ye Cai, Yue Jiang, Yang Li, and Mei Tian. 2025. Logging-data-driven lithology identification in complex reservoirs: an example from the Niuxintuo block of the Liaohe oilfield.Frontiers in Earth Science13 (2025), 1491334

2025
[19]

Shanghua Gao, Teddy Koker, Owen Queen, Tom Hartvigsen, Theodoros Tsiligkaridis, and Marinka Zitnik. 2024. Units: A unified multi-task time series model.Advances in Neural Information Processing Systems37 (2024), 140589– 140631

2024
[20]

Mononito Goswami, Konrad Szafer, Arjun Choudhry, Yifu Cai, Shuo Li, and Artur Dubrawski. 2024. MOMENT: a family of open time-series foundation models. InProceedings of the 41st International Conference on Machine Learning(Vienna, Austria)(ICML’24). JMLR.org, Article 642, 38 pages

2024
[21]

Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al . 2025. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948(2025)

work page internal anchor Pith review arXiv 2025
[22]

Michael Hüsken and Peter Stagge. 2003. Recurrent neural networks for time series classification.Neurocomputing50 (2003), 223–235

2003
[23]

Hassan Ismail Fawaz, Germain Forestier, Jonathan Weber, Lhassane Idoumghar, and Pierre-Alain Muller. 2019. Deep learning for time series classification: a review.Data mining and knowledge discovery33, 4 (2019), 917–963

2019
[24]

Hassan Ismail Fawaz, Benjamin Lucas, Germain Forestier, Charlotte Pelletier, Daniel F Schmidt, Jonathan Weber, Geoffrey I Webb, Lhassane Idoumghar, Pierre- Alain Muller, and François Petitjean. 2020. Inceptiontime: Finding alexnet for time series classification.Data Mining and Knowledge Discovery34, 6 (2020), 1936–1962

2020
[25]

Shiyi Jiang, Panke Sun, Fengqing Lyu, Sicheng Zhu, Ruifeng Zhou, Bin Li, Taihong He, Yujian Lin, Yining Gao, Wendan Song, et al. 2024. Machine learning (ML) for fluvial lithofacies identification from well logs: A hybrid classification model integrating lithofacies characteristics, logging data distributions, and ML models applicability.Geoenergy Science ...

2024
[26]

Fazle Karim, Somshubra Majumdar, Houshang Darabi, and Samuel Harford. 2019. Multivariate LSTM-FCNs for time series classification.Neural networks116 (2019), 237–245

2019
[27]

Rohit J Kate. 2016. Using dynamic time warping distances as features for improved time series classification.Data mining and knowledge discovery30 (2016), 283– 312

2016
[28]

Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu. 2017. LightGBM: a highly efficient gradient boosting decision tree. InProceedings of the 31st International Conference on Neural Information Processing Systems(Long Beach, California, USA)(NIPS’17). Curran Associates Inc., Red Hook, NY, USA, 3149–3157

2017
[29]

Diederik P Kingma. 2014. Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980(2014)

work page internal anchor Pith review arXiv 2014
[30]

Lundi Kusuma and Alwin Djamaoeddin. 2025. Leveraging Hidden Markov Model with Window Filtering Technique for Unsupervised Facies Detection in Well Logs. InSPE Middle East Oil and Gas Show and Conference. SPE, D031S085R006

2025
[31]

Zhi Li, Zhefeng Wang, Zhicheng Wei, Xiangguang Zhou, Yijun Wang, Baoxing Huai, Qi Liu, Nicholas Jing Yuan, Renbin Gong, and Enhong Chen. 2021. Cross- oilfield reservoir classification via multi-scale sensor knowledge transfer. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 35. 4215–4223

2021
[32]

Jason Lines, Sarah Taylor, and Anthony Bagnall. 2018. Time series classifica- tion with HIVE-COTE: The hierarchical vote collective of transformation-based ensembles.ACM Transactions on Knowledge Discovery from Data (TKDD)12, 5 (2018), 1–35

2018
[33]

Jiming Liu and Dongjin Xu. 2025. Logging-data-driven lithology identification of conglomerate reservoir by the assistance of integrated machine learning methods. Scientific Reports(2025)

2025
[34]

Qi Liu, Enhong Chen, Hui Xiong, Chris HQ Ding, and Jian Chen. 2011. Enhancing collaborative filtering by user interest expansion via personalized ranking.IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics)42, 1 (2011), 218–233

2011
[35]

Yuwen Liu, Yulan Zhang, Xingyuan Mao, Xucheng Zhou, Jingwen Chang, Wenwei Wang, Pan Wang, and Lianyong Qi. 2024. Lithological facies classification using attention-based gated recurrent unit.Tsinghua Science and Technology29, 4 (2024), 1206–1218

2024
[36]

Zhege Liu, Junxing Cao, Jiachun You, Shuna Chen, Yujia Lu, and Peng Zhou
[37]

A lithological sequence classification method with well log via SVM- assisted bi-directional GRU-CRF neural network.Journal of Petroleum Science and Engineering205 (2021), 108913

2021
[38]

Zhiding Liu, Jiqian Yang, Mingyue Cheng, Yucong Luo, and Zhi Li. 2024. Gen- erative Pretrained Hierarchical Transformer for Time Series Forecasting. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining(Barcelona, Spain)(KDD ’24). Association for Computing Machinery, New York, NY, USA, 2003–2013. doi:10.1145/3637528.3671855

work page doi:10.1145/3637528.3671855 2024
[39]

Yucong Luo, Yitong Zhou, Mingyue Cheng, Jiahao Wang, Daoyu Wang, Tingyue Pan, and Jintao Zhang. 2025. Time series forecasting as reasoning: A slow- thinking approach with reinforced llms.arXiv preprint arXiv:2506.10630(2025)

work page internal anchor Pith review Pith/arXiv arXiv 2025
[40]

Reiichiro Nakano, Jacob Hilton, Suchir Balaji, Jeff Wu, Long Ouyang, Christina Kim, Christopher Hesse, Shantanu Jain, Vineet Kosaraju, William Saunders, et al
[41]

Webgpt: Browser-assisted question-answering with human feedback.arXiv preprint arXiv:2112.09332(2021)

work page internal anchor Pith review arXiv 2021
[42]

An Hai Nguyen, Khang Nguyen, and Nga Mai. 2025. Classifying reservoir facies using attention-based residual neural networks.PeerJ Computer Science11 (2025), e2977

2025
[43]

Charles Packer, Vivian Fang, Shishir_G Patil, Kevin Lin, Sarah Wooders, and Joseph_E Gonzalez. 2023. MemGPT: Towards LLMs as Operating Systems. (2023)

2023
[44]

Qingwei Pang, Chenglizhao Chen, Youzhuang Sun, and Shanchen Pang. 2025. Empowering lithology identification with FreLog: Leveraging frequency domain insights in well logging signal processing.Measurement246 (2025), 116710. Conference acronym ’XX, June 03–05, 2018, Woodstock, NY Jiahao Wang, Mingyue Cheng, Yitong Zhou, Qingyang Mao, Xiaoyu Tao, Qi Liu, Enhong Chen

2025
[45]

Gyeong-Tae Park, Jina Jeong, Irina Emelyanova, Marina Pervukhina, Lionel Esteban, and Seong-Taek Yun. 2022. Data-driven sequence labeling methods incorporating the long-range spatial variation of geological data for lithofacies sequence estimation.Journal of Petroleum Science and Engineering208 (2022), 109345

2022
[46]

Pal Washa Shahzad Rathore, Matloob Hussain, Muhammad Bilal Malik, and Yawar Amin. 2023. Well log analysis and comparison of supervised machine learning algorithms for lithofacies identification in pab formation, lower indus basin.Journal of Applied Geophysics219 (2023), 105199

2023
[47]

HAN Ruiyi, WANG Zhuwen, WANG Wenhua, XU Fanghui, QI Xinghua, and CUI Yitong. 2021. Lithology identification of igneous rocks based on XGboost and conventional logging curves, a case study of the eastern depression of Liaohe Basin.Journal of Applied Geophysics195 (2021), 104480

2021
[48]

Muhammad Ansar Saleem, Ghulam Mohyuddin Sohail, Saif Ur Rehman, Qa- mar Yasin, and Ahmed E Radwan. 2025. Multiple machine learning algorithms for lithofacies prediction in the deltaic depositional system of the lower Goru Formation, Lower Indus Basin, Pakistan.Scientific Reports15, 1 (2025), 34933

2025
[49]

Timo Schick, Jane Dwivedi-Yu, Roberto Dessì, Roberta Raileanu, Maria Lomeli, Eric Hambro, Luke Zettlemoyer, Nicola Cancedda, and Thomas Scialom. 2023. Toolformer: Language models can teach themselves to use tools.Advances in Neural Information Processing Systems36 (2023), 68539–68551

2023
[50]

Agnes Schumann. 2002. Hidden Markov models for lithological well log classifi- cation.Terra Nostra4 (2002), 373–378

2002
[51]

Heng Shi, ZhenHao Xu, Peng Lin, and Wen Ma. 2023. Refined lithology identifica- tion: Methodology, challenges and prospects.Geoenergy Science and Engineering 231 (2023), 212382

2023
[52]

PAN Shijun and Shujia Qin. 2025. Application of Multimodal for Lithology-aware Prompt Engineering in Mineral Processing Systems.Intelligence, Informatics and Infrastructure6, 3 (2025), 47–54

2025
[53]

Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. 2023. Reflexion: Language agents with verbal reinforcement learning. Advances in Neural Information Processing Systems36 (2023), 8634–8652

2023
[54]

Xiaoyu Tao, Mingyue Cheng, Ze Guo, Shuo Yu, Yaguo Liu, Qi Liu, and Shijin Wang. 2026. MemCast: Memory-Driven Time Series Forecasting with Experience- Conditioned Reasoning.arXiv preprint arXiv:2602.03164(2026)

work page arXiv 2026
[55]

Xiaoyu Tao, Mingyue Cheng, Chuang Jiang, Tian Gao, Huanjian Zhang, and Yaguo Liu. 2026. Cast-R1: Learning Tool-Augmented Sequential Decision Policies for Time Series Forecasting.arXiv preprint arXiv:2602.13802(2026)

work page arXiv 2026
[56]

Hind Taud and Jean-Franccois Mas. 2017. Multilayer perceptron (MLP). In Geomatic approaches for modeling land change scenarios. Springer, 451–455

2017
[57]

Michele Tonutti, Emanuele Ruffaldi, Alessandro Cattaneo, and Carlo Alberto Avizzano. 2019. Robust and subject-independent driving manoeuvre anticipation through domain-adversarial recurrent neural networks.Robotics and Autonomous Systems115 (2019), 162–173

2019
[58]

Jiahao Wang, Mingyue Cheng, Qingyang Mao, Yitong Zhou, Daoyu Wang, Qi Liu, Feiyang Xu, and Xin Li. 2025. Tabletime: Reformulating time series classification as training-free table understanding with large language models. InProceedings of the 34th ACM International Conference on Information and Knowledge Management. 3009–3019

2025
[59]

Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. 2022. Self-consistency improves chain of thought reasoning in language models.arXiv preprint arXiv:2203.11171(2022)

work page internal anchor Pith review arXiv 2022
[60]

Zihao Wang, Shaofei Cai, Guanzhou Chen, Anji Liu, Xiaojian Shawn Ma, and Yitao Liang. 2023. Describe, explain, plan and select: interactive planning with llms enables open-world multi-task agents.Advances in Neural Information Processing Systems36 (2023), 34153–34189

2023
[61]

Zhiguang Wang, Weizhong Yan, and Tim Oates. 2016. Time series classification from scratch with deep neural networks: A strong baseline.2017 International Joint Conference on Neural Networks (IJCNN)(2016), 1578–1585. https://api. semanticscholar.org/CorpusID:14303613

2016
[62]

Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models.Advances in neural information processing systems35 (2022), 24824–24837

2022
[63]

Xuanzhi Wu and Edo Nyland. 1987. Automated stratigraphic interpretation of well-log data.Geophysics52, 12 (1987), 1665–1676

1987
[64]

Zhenhao Xu, Zhaoyang Wang, Shucai Li, Xiao Zhang, and Peng Lin. 2024. GeoPredict-LLM: Intelligent tunnel advanced geological prediction by repro- gramming large language models.Intelligent Geoengineering1, 1 (2024), 49–57

2024
[65]

Hao Xue and Flora D Salim. 2023. Promptcast: A new prompt-based learning paradigm for time series forecasting.IEEE Transactions on Knowledge and Data Engineering36, 11 (2023), 6851–6864

2023
[66]

Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. 2023. React: Synergizing reasoning and acting in language models. InInternational Conference on Learning Representations (ICLR)

2023
[67]

Le Yu, Alok Porwal, Eun-Jung Holden, and Michael C Dentith. 2012. Towards automatic lithological classification from remote sensing data using support vector machines.Computers & Geosciences45 (2012), 229–239

2012
[68]

Xiaohan Zhang, Tian Gao, Mingyue Cheng, Bokai Pan, Ze Guo, Yaguo Liu, and Xiaoyu Tao. 2025. AlphaCast: A Human Wisdom-LLM Intelligence Co- Reasoning Framework for Interactive Time Series Forecasting.arXiv preprint arXiv:2511.08947(2025)

work page internal anchor Pith review Pith/arXiv arXiv 2025
[69]

Yi Zheng, Qi Liu, Enhong Chen, Yong Ge, and J Leon Zhao. 2014. Time se- ries classification using multi-channels deep convolutional neural networks. In International conference on web-age information management. Springer, 298–310

2014
[70]

Denny Zhou, Nathanael Schärli, Le Hou, Jason Wei, Nathan Scales, Xuezhi Wang, Dale Schuurmans, Claire Cui, Olivier Bousquet, Quoc V Le, and Ed H. Chi. 2023. Least-to-Most Prompting Enables Complex Reasoning in Large Language Models. InThe Eleventh International Conference on Learning Representations. https: //openreview.net/forum?id=WZH7099tgfM

2023
[71]

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. InProceedings of the AAAI conference on artificial intelligence, Vol. 35. 11106–11115

2021
[72]

Well logs

Tian Zhou, Peisong Niu, Liang Sun, Rong Jin, et al . 2023. One fits all: Power general time series analysis by pretrained lm.Advances in neural information processing systems36 (2023), 43322–43355. GeoDecider: A Coarse-to-Fine Agentic Workflow for Explainable Lithology Classification Conference acronym ’XX, June 03–05, 2018, Woodstock, NY Table 6: Detaile...

2023