Academic Journal
Προβολή σε EDS

Toward advanced management of sludge bulking: An explainable evaluation framework to quantify and improve large language model cognitive capabilities.

Λεπτομέρειες βιβλιογραφικής εγγραφής
Τίτλος: Toward advanced management of sludge bulking: An explainable evaluation framework to quantify and improve large language model cognitive capabilities.
Συγγραφείς: Xu B; School of Technology for Sustainability, Beijing Normal University, China., Fan N; School of Technology for Sustainability, Beijing Normal University, China., Zhang C; School of Technology for Sustainability, Beijing Normal University, China., Xu G; School of Technology for Sustainability, Beijing Normal University, China., Yu H; School of Civil Engineering and Transportation, Guangzhou University, China., Zeng H; College of the Environment & Ecology, Xiamen University, China., Tong R; DeepMind, Google, 1600 Amphitheatre Pkwy, Mountain View, CA, United States., Yu K; Beijing Originwater Technology Co. Ltd., Beijing, China., Yuan Z; School of Energy and Environment, City University of Hong Kong, Hong Kong, China., Ng HY; School of Technology for Sustainability, Beijing Normal University, China; Department of Civil and Environmental Engineering, National University of Singapore, Singapore. Electronic address: huanghy@bnu.edu.cn.
Πηγή: Water research [Water Res] 2026 Oct 15; Vol. 305, pp. 126492. Date of Electronic Publication: 2026 Jul 13.
Τύπος έκδοσης: Journal Article
Γλώσσα: English
Στοιχεία περιοδικού: Publisher: Pergamon Press Country of Publication: England NLM ID: 0105072 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1879-2448 (Electronic) Linking ISSN: 00431354 NLM ISO Abbreviation: Water Res Subsets: MEDLINE
Imprint Name(s): Original Publication: Oxford, Pergamon Press.
Ιατρικοί όροι (MeSH): Waste Disposal, Fluid*/methods , Sewage*, Large Language Models
Περίληψη: Activated sludge bulking is a globally prevalent operational failure in wastewater treatment plants (WWTPs), which demands timely, accurate human intervention, but the scarcity of specialist wastewater expertise poses a considerable challenge. Large language models (LLMs) potentially support bulking management by providing human-like cognitive capabilities. However, there is no explainable evaluation framework to quantify LLMs' cognitive capabilities' strengths and limits, and, crucially, inform how these capabilities can be improved for advanced sludge bulking management. Here, we developed an explainable evaluation framework, by establishing an open-access evaluation suite (280 tasks including real-world cases) with an LLM-rater (with ground truth in prompts to ensure evaluation reliability), to diagnose LLMs' cognitive capabilities of understanding, analysis, and practice. Under this framework, we found that performance was dominated by task type rather than LLM identity (DeepSeek-R1, GPT-4o and Gemini-2.0). Concretely, LLMs generally showed limited understanding capability on microscopy-based recognition (accuracy<54%) and microbial identification (normalized-scores:28-42%), and weak analysis capability for process-microbe interactions (normalized-scores:36-52%). Practice tasks were also challenging, with deficits across all cognitive capabilities, and had highest token demand among all task types. Guided by this explainable framework, we integrated a vision module with LLMs to improve visual understanding of key filaments, including Beggiatoa, fungi, and Nocardia (accuracy from 51% to 78%). In parallel, retrieval-augmented generation (RAG) was further employed to selectively enhance practice capability, with DeepSeek-R1, a thinking model, showing the largest gains in practice capability (69%). The RAG-enhanced DeepSeek-R1 (i.e., FilamentGPT) achieved the best overall performance (normalized-scores>80% for all, except microbial identification), and demonstrated practical feasibility for WWTP decision-support based on perspectives of 20 WWTP engineers. We attribute this advantage to its stronger ability to integrate retrieved knowledge into reasonable, plant-specific workflows. Collectively, this work established an explainable evaluation framework for LLMs in sludge bulking management, constituting a science foundation for LLMs evaluation. Informed by explainable framework, we also outlined (post-)training pathways (e.g., RAG, vision module) to selectively enhance cognitive capabilities of LLMs, which could democratize subspecialist-level expertise for wastewater treatment in WWTPs.
(Copyright © 2026 Elsevier Ltd. All rights reserved.)
Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Contributed Indexing: Keywords: Filamentous microbes; Generative artificial intelligence; Inference; LLM-as-a-judge; Multimodal; RAG
Substance Nomenclature: 0 (Sewage)
Entry Date(s): Date Created: 20260723 Date Completed: 20260909 Latest Revision: 20260909
Update Code: 20260909
DOI: 10.1016/j.watres.2026.126492
PMID: 42492211
Βάση Δεδομένων: MEDLINE
FullText Text:
  Availability: 0
CustomLinks:
  – Url: https://www.doi.org/10.1016/j.watres.2026.126492?
    Name: ScienceDirect (all content) (s7799221)
    Category: fullText
    Text: View record from ScienceDirect
    MouseOverText: View record from ScienceDirect
  – Url: https://resolver.ebsco.com/c/fiv2js/result?sid=EBSCO:cmedm&genre=article&issn=18792448&ISBN=&volume=305&issue=&date=20261015&spage=126492&pages=126492&title=Water research&atitle=Toward%20advanced%20management%20of%20sludge%20bulking%3A%20An%20explainable%20evaluation%20framework%20to%20quantify%20and%20improve%20large%20language%20model%20cognitive%20capabilities.&aulast=Xu%20B&id=DOI:10.1016/j.watres.2026.126492
    Name: Full Text Finder (for New FTF UI) (ns324271)
    Category: fullText
    Text: Full Text Finder
    MouseOverText: Full Text Finder
Header DbId: cmedm
DbLabel: MEDLINE
An: 42492211
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Toward advanced management of sludge bulking: An explainable evaluation framework to quantify and improve large language model cognitive capabilities.
– Name: Author
  Label: Authors
  Group: Au
  Data: &lt;searchLink fieldCode=&quot;AU&quot; term=&quot;%22Xu+B%22&quot;&gt;Xu B&lt;/searchLink&gt;; School of Technology for Sustainability, Beijing Normal University, China.&lt;br /&gt;&lt;searchLink fieldCode=&quot;AU&quot; term=&quot;%22Fan+N%22&quot;&gt;Fan N&lt;/searchLink&gt;; School of Technology for Sustainability, Beijing Normal University, China.&lt;br /&gt;&lt;searchLink fieldCode=&quot;AU&quot; term=&quot;%22Zhang+C%22&quot;&gt;Zhang C&lt;/searchLink&gt;; School of Technology for Sustainability, Beijing Normal University, China.&lt;br /&gt;&lt;searchLink fieldCode=&quot;AU&quot; term=&quot;%22Xu+G%22&quot;&gt;Xu G&lt;/searchLink&gt;; School of Technology for Sustainability, Beijing Normal University, China.&lt;br /&gt;&lt;searchLink fieldCode=&quot;AU&quot; term=&quot;%22Yu+H%22&quot;&gt;Yu H&lt;/searchLink&gt;; School of Civil Engineering and Transportation, Guangzhou University, China.&lt;br /&gt;&lt;searchLink fieldCode=&quot;AU&quot; term=&quot;%22Zeng+H%22&quot;&gt;Zeng H&lt;/searchLink&gt;; College of the Environment &amp; Ecology, Xiamen University, China.&lt;br /&gt;&lt;searchLink fieldCode=&quot;AU&quot; term=&quot;%22Tong+R%22&quot;&gt;Tong R&lt;/searchLink&gt;; DeepMind, Google, 1600 Amphitheatre Pkwy, Mountain View, CA, United States.&lt;br /&gt;&lt;searchLink fieldCode=&quot;AU&quot; term=&quot;%22Yu+K%22&quot;&gt;Yu K&lt;/searchLink&gt;; Beijing Originwater Technology Co. Ltd., Beijing, China.&lt;br /&gt;&lt;searchLink fieldCode=&quot;AU&quot; term=&quot;%22Yuan+Z%22&quot;&gt;Yuan Z&lt;/searchLink&gt;; School of Energy and Environment, City University of Hong Kong, Hong Kong, China.&lt;br /&gt;&lt;searchLink fieldCode=&quot;AU&quot; term=&quot;%22Ng+HY%22&quot;&gt;Ng HY&lt;/searchLink&gt;; School of Technology for Sustainability, Beijing Normal University, China; Department of Civil and Environmental Engineering, National University of Singapore, Singapore. Electronic address: huanghy@bnu.edu.cn.
– Name: TitleSource
  Label: Source
  Group: Src
  Data: &lt;searchLink fieldCode=&quot;JN&quot; term=&quot;%220105072%22&quot;&gt;Water research&lt;/searchLink&gt; [Water Res] 2026 Oct 15; Vol. 305, pp. 126492. &lt;i&gt;Date of Electronic Publication: &lt;/i&gt;2026 Jul 13.
– Name: TypePub
  Label: Publication Type
  Group: TypPub
  Data: Journal Article
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: TitleSource
  Label: Journal Info
  Group: Src
  Data: &lt;i&gt;Publisher: &lt;/i&gt;&lt;searchLink fieldCode=&quot;PB&quot; term=&quot;%22Pergamon+Press%22&quot;&gt;Pergamon Press &lt;/searchLink&gt;&lt;i&gt;Country of Publication: &lt;/i&gt;England &lt;i&gt;NLM ID: &lt;/i&gt;0105072 &lt;i&gt;Publication Model: &lt;/i&gt;Print-Electronic &lt;i&gt;Cited Medium: &lt;/i&gt;Internet &lt;i&gt;ISSN: &lt;/i&gt;1879-2448 (Electronic) &lt;i&gt;Linking ISSN: &lt;/i&gt;&lt;searchLink fieldCode=&quot;IS&quot; term=&quot;%2200431354%22&quot;&gt;00431354 &lt;/searchLink&gt;&lt;i&gt;NLM ISO Abbreviation: &lt;/i&gt;Water Res &lt;i&gt;Subsets: &lt;/i&gt;MEDLINE
– Name: PublisherInfo
  Label: Imprint Name(s)
  Group: PubInfo
  Data: &lt;i&gt;Original Publication&lt;/i&gt;: Oxford, Pergamon Press.
– Name: SubjectMESH
  Label: MeSH Terms
  Group: Su
  Data: &lt;searchLink fieldCode=&quot;MM&quot; term=&quot;%22Waste+Disposal%2C+Fluid%22&quot;&gt;Waste Disposal, Fluid*&lt;/searchLink&gt;/&lt;searchLink fieldCode=&quot;MM&quot; term=&quot;%22Waste+Disposal%2C+Fluid+methods%22&quot;&gt;methods&lt;/searchLink&gt; &lt;br /&gt;&lt;searchLink fieldCode=&quot;MM&quot; term=&quot;%22Sewage%22&quot;&gt;Sewage*&lt;/searchLink&gt;&lt;br /&gt;&lt;searchLink fieldCode=&quot;MH&quot; term=&quot;%22Large+Language+Models%22&quot;&gt;Large Language Models&lt;/searchLink&gt;
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Activated sludge bulking is a globally prevalent operational failure in wastewater treatment plants (WWTPs), which demands timely, accurate human intervention, but the scarcity of specialist wastewater expertise poses a considerable challenge. Large language models (LLMs) potentially support bulking management by providing human-like cognitive capabilities. However, there is no explainable evaluation framework to quantify LLMs&#39; cognitive capabilities&#39; strengths and limits, and, crucially, inform how these capabilities can be improved for advanced sludge bulking management. Here, we developed an explainable evaluation framework, by establishing an open-access evaluation suite (280 tasks including real-world cases) with an LLM-rater (with ground truth in prompts to ensure evaluation reliability), to diagnose LLMs&#39; cognitive capabilities of understanding, analysis, and practice. Under this framework, we found that performance was dominated by task type rather than LLM identity (DeepSeek-R1, GPT-4o and Gemini-2.0). Concretely, LLMs generally showed limited understanding capability on microscopy-based recognition (accuracy&amp;lt;54%) and microbial identification (normalized-scores:28-42%), and weak analysis capability for process-microbe interactions (normalized-scores:36-52%). Practice tasks were also challenging, with deficits across all cognitive capabilities, and had highest token demand among all task types. Guided by this explainable framework, we integrated a vision module with LLMs to improve visual understanding of key filaments, including Beggiatoa, fungi, and Nocardia (accuracy from 51% to 78%). In parallel, retrieval-augmented generation (RAG) was further employed to selectively enhance practice capability, with DeepSeek-R1, a thinking model, showing the largest gains in practice capability (69%). The RAG-enhanced DeepSeek-R1 (i.e., FilamentGPT) achieved the best overall performance (normalized-scores&amp;gt;80% for all, except microbial identification), and demonstrated practical feasibility for WWTP decision-support based on perspectives of 20 WWTP engineers. We attribute this advantage to its stronger ability to integrate retrieved knowledge into reasonable, plant-specific workflows. Collectively, this work established an explainable evaluation framework for LLMs in sludge bulking management, constituting a science foundation for LLMs evaluation. Informed by explainable framework, we also outlined (post-)training pathways (e.g., RAG, vision module) to selectively enhance cognitive capabilities of LLMs, which could democratize subspecialist-level expertise for wastewater treatment in WWTPs.&lt;br /&gt; (Copyright &#169; 2026 Elsevier Ltd. All rights reserved.)
– Name: Abstract
  Label: Competing Interests
  Group: Ab
  Data: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
– Name: SubjectMinor
  Label: Contributed Indexing
  Group:
  Data: &lt;i&gt;Keywords: &lt;/i&gt;Filamentous microbes; Generative artificial intelligence; Inference; LLM-as-a-judge; Multimodal; RAG
– Name: NumberCAS
  Label: Substance Nomenclature
  Group: ID
  Data: 0 (Sewage)
– Name: DateEntry
  Label: Entry Date(s)
  Group: Date
  Data: &lt;i&gt;Date Created: &lt;/i&gt;20260723 &lt;i&gt;Date Completed: &lt;/i&gt;20260909 &lt;i&gt;Latest Revision: &lt;/i&gt;20260909
– Name: DateUpdate
  Label: Update Code
  Group: Date
  Data: 20260909
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1016/j.watres.2026.126492
– Name: AN
  Label: PMID
  Group: ID
  Data: 42492211
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=cmedm&AN=42492211
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1016/j.watres.2026.126492
    Languages:
      – Code: eng
        Text: English
    PhysicalDescription:
      Pagination:
        StartPage: 126492
    Subjects:
      – SubjectFull: Large Language Models
        Type: general
      – SubjectFull: Waste Disposal, Fluid methods
        Type: general
      – SubjectFull: Sewage
        Type: general
    Titles:
      – TitleFull: Toward advanced management of sludge bulking: An explainable evaluation framework to quantify and improve large language model cognitive capabilities.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Xu B
      – PersonEntity:
          Name:
            NameFull: Fan N
      – PersonEntity:
          Name:
            NameFull: Zhang C
      – PersonEntity:
          Name:
            NameFull: Xu G
      – PersonEntity:
          Name:
            NameFull: Yu H
      – PersonEntity:
          Name:
            NameFull: Zeng H
      – PersonEntity:
          Name:
            NameFull: Tong R
      – PersonEntity:
          Name:
            NameFull: Yu K
      – PersonEntity:
          Name:
            NameFull: Yuan Z
      – PersonEntity:
          Name:
            NameFull: Ng HY
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 15
              M: 10
              Text: 2026 Oct 15
              Type: published
              Y: 2026
          Identifiers:
            – Type: issn-electronic
              Value: 1879-2448
          Numbering:
            – Type: volume
              Value: 305
          Titles:
            – TitleFull: Water research
              Type: main
ResultId 1