Pith. sign in

REVIEW 4 major objections 5 minor 43 references

WebLists: Extracting Structured Information From Complex Interactive Websites Using Executable LLM Agents

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read BardeenAgent records a single extraction, converts it to a program with generalizable CSS selectors, and replays it to double the recall of the best prior web agent on 200 live data-extraction tasks.

desk verdict A genuinely useful benchmark and a plausible record-and-replay method, but the headline recall numbers rest on a per-agent, single-column matching rule that the authors themselves admit is forgiving. read the letter →

arxiv 2504.12682 v1 pith:5JRNNL5W submitted 2025-04-17 cs.AI cs.CL

Arth Bohra , Manvel Saroyan , Danil Melkozerov , Vahe Karufanyan , Gabriel Maher , Pascal Weinberger , Artem Harutyunyan , Giovanni Campagna This is my paper · ORCID
classification cs.AIcs.CL
keywords webagentsstructureddataextractionrecordandreplayCSSselectorslivewebsitebenchmarkLLMListsatscale
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that structured data extraction from live websites is a distinct, under-tested capability for LLM web agents, and that the right way to scale it is not to ask the LLM to extract every row. Instead, an agent should record one successful extraction on the first item of a list, convert the recorded actions into a reusable program keyed to generalizable CSS selectors, then replay that program across all items and pages. To measure this, the paper introduces WebLists, a benchmark of 200 extraction tasks on 50 live company websites across four use cases, and reports that its BardeenAgent reaches 66.2% recall, more than double the 30.5% recall of the best baseline, with 72.5% precision and roughly one-third the cost per correctly extracted row. The central claim is that the record-and-replay design, not prompt tuning, is what produces this gain.

What carries the argument

The load-bearing mechanism is the pair consisting of the EnterList tool and generalizable CSS selector generation. EnterList takes representative elements, a description, and pagination controls, identifies their least common ancestor, and scopes all subsequent actions to the first list item; the recorded actions are then replayed as a loop over the CSS-matched elements. Selector generation has two paths: a heuristic that samples diverse selector strategies and joins the survivors, and an LLM-based SelectorModel that takes the goal, column specification, the HTML of the least common parent, and a list description to produce a selector for irregular layouts. This machinery is what lets the agent generalize from one item to the whole dataset without additional LLM calls per row.

What would settle it

Create a controlled variant of WebLists pages in which each list item is wrapped in a different parent element with no shared selector, run BardeenAgent's record-replay pipeline, and check whether recall collapses from 66% toward the 30.5% baseline; if recall stays high, the shared-selector assumption is not what carries the gain.

Watch

Extended reading notes

Core claim

The central discovery is that the regular structure of HTML can be turned into an executable extraction program. In the record phase, the LLM navigates to the page, uses the EnterList tool to select representative list elements, and records operations on only the first list item while all CSS selectors are scoped to that item. In the replay phase, those operations become a loop over every element matching a generalizable selector, with pagination handling, duplicate suppression, and a two-pass strategy that first collects item URLs and only then visits each one. Because LLM calls happen mainly during recording, the number of LLM calls does not grow with dataset size, which the paper credits for its recall gain and for the 3x reduction in cost per extracted row.

Load-bearing premise

The replay phase assumes the target data is exposed as repeated DOM nodes that share a generalizable CSS selector; if a website's list items have heterogeneous markup, or the data only appears after unique per-item interaction, the recorded selectors over-match or under-match and recall collapses.

Editorial extensions

If this is right

  • If the record-and-replay design is correct, large-scale extraction no longer needs one LLM call per item; the LLM decides how to extract once, and the loop carries out the extraction.
  • WebLists can be re-run at any time on the same live websites because its evaluation scripts regenerate ground truth, making deterministic live-web evaluation possible without simulator proxies.
  • Agents optimized for navigation and question answering are poor at structured extraction, so benchmarks that only test those skills will miss a failure mode that matters for business use.
  • The gap between the Jobs use case (86.8% recall) and the Job Categories use case (64.5% recall) shows that filtering and multi-category list structure, not basic pagination, are the main remaining recall bottleneck.

Reading between the lines

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

  • Editorial inference: If the record-and-replay pattern generalizes, the same trick should work for any repeating DOM structure — tables, grids, card walls, and paginated search results — though the paper only evaluates lists.
  • Editorial inference: Because the AnswerQuestion tool still invokes an LLM on every replayed item, the cost-scaling advantage is strongest when the schema can be satisfied by direct extraction or regex; strictly per-item QA schemas would partially erode the reported 3x cost reduction.
  • Editorial inference: A stress test the paper does not run is to mutate list markup between pagination pages, such as changing class names after page 2; a selector derived from the first page alone would likely miss later items and reveal how much of the gain depends on markup homogeneity.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper introduces WebLists, a benchmark of 200 structured data-extraction tasks across 50 live websites and four use cases (jobs, job categories, blogs, testimonials), together with BardeenAgent, a two-phase LLM web agent that records actions as executable programs with CSS selectors and replays them to extract large lists. On the benchmark, BardeenAgent reports 66.2% recall and 72.5% precision, roughly doubling the recall of the strongest baseline Wilbur, and reducing cost per extracted row by about 3x. The paper also includes a question-answering evaluation on 50 questions and an ablation of the LLM-based selector model.

Significance. If the reported results hold, the paper makes a useful contribution: WebLists is a live, deterministic benchmark with a realistic scale (200 tasks, 50 websites) that addresses a real gap in web-agent evaluation, and the record-and-replay idea with generalizable CSS selectors is a sensible and potentially efficient way to scale extraction without per-item LLM calls. The ablation without the selector model (recall drops from 66.2% to 35.6%) shows that the proposed component is important, and the explicit discussion of limitations in Section 4.6 is commendable. However, the quantitative headline claims currently depend on an evaluation rule that is not held constant across agents and that is weaker than the benchmark's own schema definition, so the size of the reported advantage is not yet established.

major comments (4)
  1. [Section 4.2, Table 3] The evaluation uses a different matching column for each agent: "We choose different columns for different agents, in order to maximize matching between gold and extracted data" (Section 4.2). Since the schema identifies rows by URL (Section 2.2) and the benchmark's task definition requires adherence to the full schema, a row that matches on only one hand-picked column may be counted as correct even if the URL identifier and other columns are wrong or missing. Section 4.6 concedes that this is a "forgiving metric." The headline comparison (66.2% vs. 30.5% recall) is therefore not computed under a fixed, task-aligned match rule. Please report results under a fixed matching column (e.g., the URL column) and under full-schema matching for all agents.
  2. [Table 4] The per-row cost advantage ("3x lower cost per correct output row") is computed using the same relaxed one-column matches as the recall figures, so the denominator of "correct rows" is inflated by the same per-agent column selection. The cost claim should be recomputed under the fixed matching rule proposed above, or the dependence of the cost ratio on the match rule should be explicitly reported.
  3. [Sections 4.2, 4.3 and Appendix C] All results are single runs with no variance or significance measures. The policy model uses temperature 0.7 (Appendix C), and live websites change over time, so the numbers in Table 3 may not be stable. Please run each configuration multiple times, or at least report per-website variance and confidence intervals, and state the temporal window of the evaluation.
  4. [Section 4.2] "We tune goals per use-case and model" and manually select few-shot examples for BardeenAgent and Wilbur. This means the comparison is not a fixed-prompt comparison; the reported improvement could be partly due to prompt engineering rather than the executable-agent mechanism. Please provide the exact prompts and few-shot examples for all agents, and ideally include an ablation in which BardeenAgent uses the same goal/schema phrasing as the baselines, to isolate the contribution of the record-and-replay mechanism.
minor comments (5)
  1. [Section 4.6] The heading "Balacing Extraction & Interaction" contains a typo and should read "Balancing Extraction & Interaction."
  2. [Appendix D] The sentence "In Fig. 3 We also show the the time for Operator to complete a request end-to-end" contains a duplicated "the" and should be rewritten, e.g., "Figure 3 also shows the time for Operator to complete a request end-to-end."
  3. [Section 1.1] The phrase "yielding an improvement of 36% in recall" is ambiguous: it refers to a 35.7 percentage-point improvement, not a 36% relative improvement. Please state this as percentage points or as a relative improvement explicitly.
  4. [Section 2.2] The benchmark limits extraction to at most 5 pages of data per task; this is an important constraint for interpreting the recall values, but it is only mentioned once. Please restate it in the evaluation section and discuss how it affects comparability with agents that may paginate further.
  5. [References] The reference to "Sams, 2025" is a blog with no formal author or venue; consider using a more standard source for web statistics, or clearly mark it as a web resource.

Circularity Check

0 steps flagged · score 2.0 of 10

No load-bearing circularity; the only issue is a minor self-citation (Wilbur baseline) plus a disclosed evaluation-metric caveat.

full rationale

The paper is a benchmark plus an empirical system paper rather than a derivation, and the main claims are measured on live websites against hand-built reference scripts. BardeenAgent is not trained on WebLists gold labels, and its record/replay/selector-generation pipeline is not constructed from the benchmark's ground truth, so the core '66% recall' result is not baked in by definition. The closest issue is a self-citation: Wilbur, the strongest baseline and the source of the verifier design in Section 3.1, is the authors' own prior system (Lutz et al., 2024). That is a real conflict of interest and a reason to scrutinize the comparison, but it is not load-bearing circularity because Wilbur's numbers are measured directly in this paper rather than assumed from the citation. Section 4.2 also discloses that matching is done on a single per-agent chosen column, and Section 4.6 concedes that 'a stricter evaluation would yield lower recall'; this is an evaluation-validity caveat that should be weighed in review, but it is stated in the paper and does not make the reported numbers equivalent to the method's inputs by construction. No circular step rises to the level of a fitted input renamed as a prediction, so the circularity score is low.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central method introduces no new physical entities. Its free parameters are evaluation and prompt choices. The strongest assumption is structural regularity of the web DOM, which the method exploits but does not prove.

free parameters (5)
  • Task goals tuned per use-case and per model = not reported
    Section 4.2 says goals were tuned for each model; prompt differences can drive web-agent performance, so this is a hand-chosen input to the comparison.
  • Few-shot examples for BardeenAgent and Wilbur = manually selected
    Section 4.2 says few-shot examples were manually selected rather than using Wilbur's auto-curriculum; this could advantage the authors' own system.
  • Matching column for precision and recall = different per agent
    Section 4.2: 'We choose different columns for different agents, in order to maximize matching between gold and extracted data.' This is a per-agent fit to the evaluation criterion.
  • Maximum pages extracted per task = 5
    Section 2.2 caps extraction at 5 pages, which bounds recall and hides failures on longer lists.
  • LLM hyperparameters = GPT-4 Turbo, GPT-4o, Gemini Flash 2.0, Gemini 1.5 Pro with listed temperatures
    Appendix C lists model and temperature choices; these are chosen by hand and affect cost and accuracy.
assumptions (4)
  • domain assumption Reference scripts produce correct, up-to-date ground truth for each live website.
    Section 2.2: evaluation runs the reference script first and treats its output as gold; annotation errors or script failures would directly corrupt the benchmark.
  • domain assumption The target data appears as repeated DOM nodes with a shared, generalizable CSS selector.
    Sections 3.2 and 3.3: EnterList and replay rely on least common ancestor and selector generalization; heterogeneous or dynamically generated markup breaks this.
  • domain assumption Exact match on one specified column is sufficient to judge that an extracted row is correct.
    Section 4.2 defines precision and recall via exact matching on a single column; the authors note a stricter evaluation would lower recall.
  • domain assumption Text-only LLMs with simplified DOM observations can make correct navigation and selector decisions.
    Appendix C uses text-only models; visual layout, which matters for widgets and forms, is not observed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of WebLists: Extracting Structured Information From Complex Interactive Websites Using Executable LLM Agents." pith.science (2026). https://pith.science/paper/5JRNNL5W

@misc{pith2026250412682,
  author       = {Pith},
  title        = {Pith review of: WebLists: Extracting Structured Information From Complex Interactive Websites Using Executable LLM Agents},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5JRNNL5W}},
  note         = {Machine review of arXiv:2504.12682}
}
read the original abstract

Most recent web agent research has focused on navigation and transaction tasks, with little emphasis on extracting structured data at scale. We present WebLists, a benchmark of 200 data-extraction tasks across four common business and enterprise use-cases. Each task requires an agent to navigate to a webpage, configure it appropriately, and extract complete datasets with well-defined schemas. We show that both LLMs with search capabilities and SOTA web agents struggle with these tasks, with a recall of 3% and 31%, respectively, despite higher performance on question-answering tasks. To address this challenge, we propose BardeenAgent, a novel framework that enables web agents to convert their execution into repeatable programs, and replay them at scale across pages with similar structure. BardeenAgent is also the first LLM agent to take advantage of the regular structure of HTML. In particular BardeenAgent constructs a generalizable CSS selector to capture all relevant items on the page, then fits the operations to extract the data. On the WebLists benchmark, BardeenAgent achieves 66% recall overall, more than doubling the performance of SOTA web agents, and reducing cost per output row by 3x.

Figures

Figures reproduced from arXiv: 2504.12682 by the authors.

Figure 1
Figure 1. BardeenAgent first records all actions it performs on the web page, then converts [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Example of BardeenAgent extracting data from a blog, with pagination. The policy [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. Completion times for OpenAI Operator agent on the WebLists Jobs use case. [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 11 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    Agent-e: From autonomous web navigation to foundational design principles in agentic systems, 2024

    Tamer Abuelsaad, Deepak Akkil, Prasenjit Dey, Ashish Jagmohan, Aditya Vempaty, and Ravi Kokku. Agent-e: From autonomous web navigation to foundational design principles in agentic systems, 2024. URL https://arxiv.org/abs/2407.13032

    arXiv 2024

  3. [3]

    Agent s: An open agentic framework that uses computers like a human, 2024

    Saaket Agashe, Jiuzhou Han, Shuyu Gan, Jiachen Yang, Ang Li, and Xin Eric Wang. Agent s: An open agentic framework that uses computers like a human, 2024. URL https://arxiv.org/abs/2410.08164

    arXiv 2024

  4. [4]

    Introducing computer use, a new claude 3.5 sonnet, and claude 3.5 haiku

    Anthropic. Introducing computer use, a new claude 3.5 sonnet, and claude 3.5 haiku. https://www.anthropic.com/news/3-5-models-and-computer-use, October 2024. URL https://www.anthropic.com/news/3-5-models-and-computer-use. Blog post

    work page 2024

  5. [5]

    Trafilatura: A web scraping library and command-line tool for text discovery and extraction

    Adrien Barbaresi. Trafilatura: A web scraping library and command-line tool for text discovery and extraction. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing: System Demonstrations, pp.\ 122--131, Online, August 2021. Association for Comp...

    doi:10.18653/v1/2021.acl-demo.15 2021

  6. [6]

    Multilingual Attribute Extraction from News Web Pages

    Pavel Bedrin, Maksim Varlamov, and Alexander Yatskov. Multilingual attribute extraction from news web pages, 2025. URL https://arxiv.org/abs/2502.02167

    work page Pith review arXiv 2025

  7. [7]

    An empirical comparison of web content extraction algorithms

    Janek Bevendorff, Sanket Gupta, Johannes Kiesel, and Benno Stein. An empirical comparison of web content extraction algorithms. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR '23, pp.\ 2594–2603, New York, NY, USA, 2023. Association for Computing Machinery. ISBN 9781450394086. doi:...

    arXiv 2023

  8. [8]

    Kulvinder Singh

    Vedant Bisht, Renu Choyal, Akshay Singh Negi, and Er. Kulvinder Singh. Utilizing python for web scraping and incremental data extraction. In 2023 2nd International Conference on Automation, Computing and Renewable Systems (ICACRS), pp.\ 1450--1455, 2023. doi:10.1109/ICACRS58579.2023.10404702

    arXiv 2023

Show all 43 references
  1. [9]

    Large language models empowered personalized web agents, 2024

    Hongru Cai, Yongqi Li, Wenjie Wang, Fengbin Zhu, Xiaoyu Shen, Wenjie Li, and Tat-Seng Chua. Large language models empowered personalized web agents, 2024. URL https://arxiv.org/abs/2410.17236

    2024 arXiv

  2. [10]

    The cloud 100 2024

    Kenrick Cai and Alex Konrad. The cloud 100 2024. https://www.forbes.com/lists/cloud100/, August 2024

    2024

  3. [11]

    Fundus: A simple-to-use news scraper optimized for high quality extractions, 2024

    Max Dallabetta, Conrad Dobberstein, Adrian Breiding, and Alan Akbik. Fundus: A simple-to-use news scraper optimized for high quality extractions, 2024. URL https://arxiv.org/abs/2403.15279

    2024 arXiv

  4. [12]

    Mind2web: Towards a generalist agent for the web, 2023

    Xiang Deng, Yu Gu, Boyuan Zheng, Shijie Chen, Samuel Stevens, Boshi Wang, Huan Sun, and Yu Su. Mind2web: Towards a generalist agent for the web, 2023. URL https://arxiv.org/abs/2306.06070

    2023 arXiv

  5. [13]

    O mer Faruk Ulusoy, Bilal Macit, Asl Ece K rkaya, \

    Y lmaz Dikilita s , C o s kun C akal, Ahmet Can Okumu s , Halime Nur Yal c n, Emine Y ld r m, \"O mer Faruk Ulusoy, Bilal Macit, Asl Ece K rkaya, \"O zkan Yal c n, Ekin Erdo g mu s , and Ahmet Sayar. Performance analysis for web scraping tools: Case studies on beautifulsoup, s...

    2024

  6. [14]

    Gemini: A family of highly capable multimodal models, 2024

    Gemini Team . Gemini: A family of highly capable multimodal models, 2024. URL https://arxiv.org/abs/2312.11805

    2024 arXiv

  7. [15]

    Is your llm secretly a world model of the internet? model-based planning for web agents, 2024

    Yu Gu, Boyuan Zheng, Boyu Gou, Kai Zhang, Cheng Chang, Sanjari Srivastava, Yanan Xie, Peng Qi, Huan Sun, and Yu Su. Is your llm secretly a world model of the internet? model-based planning for web agents, 2024. URL https://arxiv.org/abs/2411.06559

    2024 arXiv

  8. [16]

    Understanding html with large language models

    Izzeddin Gur, Ofir Nachum, Yingjie Miao, Mustafa Safdari, Austin Huang, Aakanksha Chowdhery, Sharan Narang, Noah Fiedel, and Aleksandra Faust. Understanding html with large language models. arXiv preprint arXiv:2210.03945, 2022

    arXiv

  9. [17]

    Openwebvoyager: Building multimodal web agents via iterative real-world exploration, feedback and optimization, 2024

    Hongliang He, Wenlin Yao, Kaixin Ma, Wenhao Yu, Hongming Zhang, Tianqing Fang, Zhenzhong Lan, and Dong Yu. Openwebvoyager: Building multimodal web agents via iterative real-world exploration, feedback and optimization, 2024. URL https://arxiv.org/abs/2410.19609

    2024 arXiv

  10. [18]

    A webextension framework for experimentation and evaluation of webpage segmentation methods

    Geunseong Jung and Jaehyuk Cha. A webextension framework for experimentation and evaluation of webpage segmentation methods. SoftwareX, 23: 0 101501, 2023. ISSN 2352-7110. doi:https://doi.org/10.1016/j.softx.2023.101501. URL https://www.sciencedirect.com/science/article/pii/S2...

    2023

  11. [19]

    Extracting the main content of web pages using the first impression area

    Geunseong Jung, Sungjae Han, Hansung Kim, Kwanguk Kim, and Jaehyuk Cha. Extracting the main content of web pages using the first impression area. IEEE Access, 10: 0 129958--129969, 2022. doi:10.1109/ACCESS.2022.3229080

    2022

  12. [20]

    Rap: Retrieval-augmented planning with contextual memory for multimodal llm agents, 2024

    Tomoyuki Kagaya, Thong Jing Yuan, Yuxuan Lou, Jayashree Karlekar, Sugiri Pranata, Akira Kinose, Koki Oguri, Felix Wick, and Yang You. Rap: Retrieval-augmented planning with contextual memory for multimodal llm agents, 2024

    2024

  13. [21]

    Language models can solve computer tasks

    Geunwoo Kim, Pierre Baldi, and Stephen McAleer. Language models can solve computer tasks. Advances in Neural Information Processing Systems, 36, 2024

    2024

  14. [22]

    Wilbur: Adaptive in-context learning for robust and accurate web agents, 2024

    Michael Lutz, Arth Bohra, Manvel Saroyan, Artem Harutyunyan, and Giovanni Campagna. Wilbur: Adaptive in-context learning for robust and accurate web agents, 2024. URL https://arxiv.org/abs/2404.05902

    2024 arXiv

  15. [23]

    Gpt-4 technical report, 2024

    OpenAI. Gpt-4 technical report, 2024. URL https://arxiv.org/abs/2303.08774

    2024 arXiv

  16. [24]

    Introducing operator

    OpenAI. Introducing operator. https://openai.com/index/introducing-operator/, January 2025. URL https://openai.com/index/introducing-operator/. Blog post

    2025

  17. [25]

    Webcanvas: Benchmarking web agents in online environments, 2024

    Yichen Pan, Dehan Kong, Sida Zhou, Cheng Cui, Yifei Leng, Bing Jiang, Hangyu Liu, Yanyi Shang, Shuyan Zhou, Tongshuang Wu, and Zhengyang Wu. Webcanvas: Benchmarking web agents in online environments, 2024. URL https://arxiv.org/abs/2406.12373

    2024 arXiv

  18. [26]

    Meet new sonar: A blazing fast model optimized for perplexity search

    Perplexity Team . Meet new sonar: A blazing fast model optimized for perplexity search. https://www.perplexity.ai/hub/blog/meet-new-sonar, February 2025

    2025

  19. [27]

    Agent q: Advanced reasoning and learning for autonomous ai agents, 2024

    Pranav Putta, Edmund Mills, Naman Garg, Sumeet Motwani, Chelsea Finn, Divyansh Garg, and Rafael Rafailov. Agent q: Advanced reasoning and learning for autonomous ai agents, 2024. URL https://arxiv.org/abs/2408.07199

    2024 arXiv

  20. [28]

    Webrl: Training llm web agents via self-evolving online curriculum reinforcement learning, 2025

    Zehan Qi, Xiao Liu, Iat Long Iong, Hanyu Lai, Xueqiao Sun, Wenyi Zhao, Yu Yang, Xinyue Yang, Jiadai Sun, Shuntian Yao, Tianjie Zhang, Wei Xu, Jie Tang, and Yuxiao Dong. Webrl: Training llm web agents via self-evolving online curriculum reinforcement learning, 2025. URL https:/...

    2025 arXiv

  21. [29]

    How many websites are there in the world (2025 stats)

    Daniel Sams. How many websites are there in the world (2025 stats). https://curatelabs.co/how-many-websites-are-there/, March 2025

    2025

  22. [30]

    Validation and extraction of reliable information through automated scraping and natural language inference

    Arjun Shah, Hetansh Shah, Vedica Bafna, Charmi Khandor, and Sindhu Nair. Validation and extraction of reliable information through automated scraping and natural language inference. Engineering Applications of Artificial Intelligence, 147: 0 110284, 2025. ISSN 0952-1976. doi:h...

    2025

  23. [31]

    Bearcubs: A benchmark for computer-using web agents, 2025

    Yixiao Song, Katherine Thai, Chau Minh Pham, Yapei Chang, Mazin Nadaf, and Mohit Iyyer. Bearcubs: A benchmark for computer-using web agents, 2025. URL https://arxiv.org/abs/2503.07919

    2025 arXiv

  24. [32]

    Dom based content extraction via text density

    Fei Sun, Dandan Song, and Lejian Liao. Dom based content extraction via text density. Proceedings of the 34th international ACM SIGIR conference on Research and development in Information Retrieval, 2011. URL https://api.semanticscholar.org/CorpusID:10355129

    2011

  25. [33]

    Chan, Jikun Kang, Wenqi Wu, Filippos Christianos, Fraser Greenlee, Andy Toulis, and Marvin Purtorab

    George Thomas, Alex J. Chan, Jikun Kang, Wenqi Wu, Filippos Christianos, Fraser Greenlee, Andy Toulis, and Marvin Purtorab. Webgames: Challenging general-purpose web-browsing ai agents, 2025. URL https://arxiv.org/abs/2502.18356

    2025 arXiv

  26. [34]

    Web2text: Deep structured boilerplate removal

    Thijs Vogels, Octavian-Eugen Ganea, and Carsten Eickhoff. Web2text: Deep structured boilerplate removal. In Gabriella Pasi, Benjamin Piwowarski, Leif Azzopardi, and Allan Hanbury (eds.), Advances in Information Retrieval, pp.\ 167--179, Cham, 2018. Springer International Publi...

    2018

  27. [35]

    Automating web data collection: Challenges, solutions, and python-based strategies for effective web scraping

    Mutaz Abdel Wahed, Mowafaq Salem Alzboon, Muhyeeddin Alqaraleh, Jaradat Ayman, Mohammad Al-Batah, and Ahmad Fuad Bader. Automating web data collection: Challenges, solutions, and python-based strategies for effective web scraping. In 2024 7th International Conference on Intern...

    2024

  28. [36]

    Tur[k]ingbench: A challenge benchmark for web agents, 2025

    Kevin Xu, Yeganeh Kordi, Tanay Nayak, Adi Asija, Yizhong Wang, Kate Sanders, Adam Byerly, Jingyu Zhang, Benjamin Van Durme, and Daniel Khashabi. Tur[k]ingbench: A challenge benchmark for web agents, 2025. URL https://arxiv.org/abs/2403.11905

    2025 arXiv

  29. [37]

    Agentoccam: A simple yet strong baseline for llm-based web agents, 2024

    Ke Yang, Yao Liu, Sapana Chaudhary, Rasool Fakoor, Pratik Chaudhari, George Karypis, and Huzefa Rangwala. Agentoccam: A simple yet strong baseline for llm-based web agents, 2024. URL https://arxiv.org/abs/2410.13825

    2024 arXiv

  30. [38]

    Webshop: Towards scalable real-world web interaction with grounded language agents

    Shunyu Yao, Howard Chen, John Yang, and Karthik Narasimhan. Webshop: Towards scalable real-world web interaction with grounded language agents. In S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh (eds.), Advances in Neural Information Processing Systems, volum...

    2022

  31. [39]

    Eliminating noisy information in web pages for data mining

    Lan Yi, Bing Liu, and Xiaoli Li. Eliminating noisy information in web pages for data mining. In Proceedings of the Ninth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD '03, pp.\ 296–305, New York, NY, USA, 2003. Association for Computing Machin...

    2003

  32. [40]

    Xu, Hao Zhu, Xuhui Zhou, Robert Lo, Abishek Sridhar, Xianyi Cheng, Tianyue Ou, Yonatan Bisk, Daniel Fried, Uri Alon, and Graham Neubig

    Shuyan Zhou, Frank F. Xu, Hao Zhu, Xuhui Zhou, Robert Lo, Abishek Sridhar, Xianyi Cheng, Tianyue Ou, Yonatan Bisk, Daniel Fried, Uri Alon, and Graham Neubig. Webarena: A realistic web environment for building autonomous agents, 2024. URL https://arxiv.org/abs/2307.13854

    2024 arXiv

  33. [41]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  34. [42]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  35. [43]

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.