mlsanityData to 7 Oct 2026

Simulator-Grounded Large Language Models for Industrial Causal Reasoning: Tool-Use, Structured Injection, and Plant-Portable Retrieval for Wastewater Treatment Decision Support

arXiv 2608.05151 ↗First posted 20 May 2026cs.CL cs.AIcited by 0

On the map: LLM reasoning → Causal reasoning in LLMs

Abstract (arXiv)

Wastewater operators need answers grounded in how their plant's variables interact and how fast effects propagate, not in generic pretraining text, when asking causal questions such as "why is N2O rising?" or "what happens if I cut aeration by 20%?". We compare three concrete ways to ground a frozen Qwen2.5-32B-Instruct model in an architecturally interpretable wastewater simulator (CCSS-IX): a live simulator oracle (Method 1), structured parameter injection (Method 2), and a Decoupled Recall-Reasoning (DRR) retriever (Method 3). On a 198-question causal benchmark the three reach 99.5%, 79%, and 75.8%, forming a deployment ladder above the strongest retrieval-augmented baseline at 48%. The DRR retriever has 110M parameters and trains per plant in ~17 seconds; after cross-plant transfer to a biologically distinct plant it still reaches 88%, while Method 2's static table cannot transfer. On a 60-question counterfactual benchmark only Method 3 handles queries about what happens after an intervention: +16.3 pp over Method 2, paired 95% CI [+7.1, +26.4] pp, with 100% on the timescale and operating-regime categories. On the AI2 Reasoning Challenge (ARC) with an OpenBookQA fact corpus, the same selective-retrieval mechanism reaches 79% versus unconstrained Llama-3.1-8B 76% and full-injection 74%, a +3 pp out-of-domain replication that argues against a result specific to wastewater treatment. We provide the first single-simulator comparison of live tool-use, static parameter injection, and learned numerical-parameter retrieval for industrial causal question answering.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Causal reasoning in LLMs, one of 17 directions in LLM reasoning. Builds on 30 of its 99 references that are in the map. First posted on arXiv 20 May 2026.

Builds on

30 of 99 references.

References · 30

  1. 2023GPT-4 Technical Reportcited by 18,132
  2. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  3. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,146
  4. 2024Qwen2.5 Technical Reportcited by 4,713
  5. 2018Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challengecited by 4,710
  6. 2022Constitutional AI: Harmlessness from AI Feedbackcited by 2,950
  7. 2023Code Llama: Open Foundation Models for Codecited by 2,105
  8. 2023PaLM 2 Technical Reportcited by 1,154
  9. 2023REPLUG: Retrieval-Augmented Black-Box Language Modelscited by 786
  10. 2022Galactica: A Large Language Model for Sciencecited by 736
  11. 2023An Empirical Study of Catastrophic Forgetting in Large Language Models During Continual Fine-tuningcited by 622
  12. 2023Textbooks Are All You Needcited by 561
18 more
  1. 2024LoRA Learns Less and Forgets Lesscited by 313
  2. 2023MEDITRON-70B: Scaling Medical Pretraining for Large Language Modelscited by 272
  3. 2024TimeXer: Empowering Transformers for Time Series Forecasting with Exogenous Variablescited by 241
  4. 2023TEST: Text Prototype Aligned Embedding to Activate LLM's Ability for Time Seriescited by 165
  5. 2023Understanding Catastrophic Forgetting in Language Models via Implicit Inferencecited by 138
  6. 2022Internet-augmented language models through few-shot prompting for open-domain question answeringcited by 131
  7. 2023Chatlaw: A Multi-Agent Legal Assistant based on a Role-Aligned Mixture-of-Experts Architecturecited by 119
  8. 2023Multiple Physics Pretraining for Physical Surrogate Modelscited by 90
  9. 2025Text-to-LoRA: Instant Transformer Adaptioncited by 50
  10. 2025Autonomous Control Leveraging LLMs: An Agentic Framework for Next-Generation Industrial Automationcited by 8
  11. 2025Knowledge Graph Fusion with Large Language Models for Accurate, Explainable Manufacturing Process Planningcited by 3
  12. 2026SimulCost: A Cost-Aware Benchmark and Toolkit for Automating Physics Simulations with LLMscited by 3
  13. 2025Large Language Models for Water Distribution Systems Modeling and Decision-Makingcited by 2
  14. 2026Systematic Integration of Digital Twins and Constrained LLMs for Interpretable Cyber-Physical Anomaly Detectioncited by 2
  15. 2026Data-Driven Open-Loop Simulation for Digital-Twin Operator Decision Support in Wastewater Treatmentcited by 2
  16. 2026Brainstacks: Cross-Domain Cognitive Capabilities via Frozen MoE-LoRA Stacks for Continual LLM Learningcited by 1
  17. 2026WaterAdmin: Orchestrating Community Water Distribution Optimization via AI Agentscited by 1
  18. 2026Explainable Wastewater Digital Twins: Adaptive Context-Conditioned Structured Simulators with Self-Falsifying Decision Supportcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026WuYuEval: A Multi-Level Benchmark for Large Language Models in Solid Waste Management
  2. 2025Bridging AI and Carbon Capture: A Dataset for LLMs in Ionic Liquids and CBE Research
  3. 2025SimulRAG: Simulator-based RAG for Grounding LLMs in Long-form Scientific QA
  4. 2026LLM Agents Perform Controlled Experiments Using Simulation Models
  5. 2024Towards a Benchmark for Causal Business Process Reasoning with LLMs
  6. 2025Simulation Agent: A Framework for Integrating Simulation and Large Language Models for Enhanced Decision-Making
  7. 2024Language Agents Meet Causality -- Bridging LLMs and Causal World Models
  8. 2025AI-Supported Platform for System Monitoring and Decision-Making in Nuclear Waste Management with Large Language Models ↗