A Neurosymbolic Approach for Constructing Planning Domain Models from Clinical Narratives

Ranveer Singh; Saurabh Mathur; Michael Skinner; Prasad Tadepalli; Kristian Kersting; Sriraam Natarajan

cs.LGAug 21, 2026

A Neurosymbolic Approach for Constructing Planning Domain Models from Clinical Narratives

Authors: Ranveer Singh, Saurabh Mathur, Michael Skinner, Prasad Tadepalli, Kristian Kersting, Sriraam Natarajan

Organizations: The University of Texas at Dallas, Richardson, TX, USA · Technical University of Darmstadt, Darmstadt, Germany · Oregon State University, Corvallis, OR, USA · Hessian Center for Artificial Intelligence (hessian.ai), Darmstadt, Germany ·

Abstract

Surgical procedures such as laparoscopic appendectomy are complex, high-stakes processes, yet formalizing their workflows for decision support remains a significant challenge. Inducing probabilistic planning domain models in this setting is particularly difficult due to the lack of structured event data and the prevalence of implicit actions in clinical narratives, which neither empirical symbolic methods nor Large Language Models (LLMs) can adequately address on their own. We introduce NSPIN, a neurosymbolic framework for inducing probabilistic planning domain models from unstructured clinical narratives. Our method extracts and imputes structured event sequences from raw text using a pretrained LLM, then induces a PPDDL model and refines its preconditions with LLM-proposed revisions, guided by empirical validation. We evaluate the approach on 2,660 laparoscopic appendectomy notes written by 9 surgeons. NSPIN yields models that generalize to unseen notes, and expert clinical review indicates its induced knowledge is largely consistent with surgical practice.

Figures & tables

Panel of Figure 1 — Figure 1: The NSPIN framework for inducing probabilistic planning models from clinical narratives. Our neurosymbolic approach extracts structured predicates from unstructured surgical notes using an LLM. It uses this structured information to synthesize a planning model that generalizes across diverse surgical workflows. Using an LLM to ground each action and observation in the text allows NSPIN to account for both explicit and implicitly mentioned information.

Figure 2: A partial surgical sequence sampled using an action-transition model and NSPIN. NSPIN learns action preconditions that constrain the set of possible next actions, along with observational effects each action probabilistically produces. These emitted observations provide context for a simple data-driven action-transition model used for sampling candidate actions. Blue nodes represent sampled actions, gray nodes represent emitted observations, orange arrows indicate candidate precondition relationships, and gray arrows indicate observation flow. In this example, port placement is followed by the observation of adhesions, which then provides contributing context for performing an adhesiolysis.

Figure 3: Surgical Note lengths and counts across surgeons

Dataset	Top-1 Acc	Top-3 Acc	Avg NLL
	$\uparrow$ is better	$\uparrow$ is better	$\downarrow$ is better
Imputed	$\mathbf{0.32\scriptstyle{\pm 0.07}}$	$\mathbf{0.51\scriptstyle{\pm 0.10}}$	$\mathbf{9.33\scriptstyle{\pm 2.35}}$
Non-Imputed	$0.24\scriptstyle{\pm 0.09}$	$0.39\scriptstyle{\pm 0.12}$	$11.85\scriptstyle{\pm 2.85}$

Table 1: Action set prediction performance comparison across imputed and non-imputed datasets

Method	Top-1 Acc	Top-3 Acc	Avg NLL
	$\uparrow$ is better	$\uparrow$ is better	$\downarrow$ is better
NSPIN (w/wo) Refinement	$\mathbf{0.88\scriptstyle{\pm 0.03}}$	$\mathbf{0.93\scriptstyle{\pm 0.03}}$	$\mathbf{1.67\scriptstyle{\pm 0.60}}$
LLM with Example	$0.56\scriptstyle{\pm 0.23}$	$0.56\scriptstyle{\pm 0.23}$	$9.22\scriptstyle{\pm 4.76}$
LLM only	$0.78\scriptstyle{\pm 0.09}$	$0.78\scriptstyle{\pm 0.09}$	$4.72\scriptstyle{\pm 1.80}$

Table 2: Observation prediction performance comparison on the imputed dataset across different methods

Method	Top-1 Acc	Top-3 Acc	Avg NLL	TADR
	$\uparrow$ is better	$\uparrow$ is better	$\downarrow$ is better	$\downarrow$ is better
NSPIN	$\mathbf{0.42\scriptstyle{\pm 0.07}}$	$\mathbf{0.64\scriptstyle{\pm 0.08}}$	$\mathbf{6.08\scriptstyle{\pm 1.57}}$	$\mathbf{0.16\scriptstyle{\pm 0.06}}$
NSPIN w/o Refinement	$0.32\scriptstyle{\pm 0.07}$	$0.51\scriptstyle{\pm 0.09}$	$9.33\scriptstyle{\pm 2.22}$	$0.36\scriptstyle{\pm 0.12}$
LLM with Example	$0.09\scriptstyle{\pm 0.16}$	$0.19\scriptstyle{\pm 0.27}$	$16.28\scriptstyle{\pm 6.34}$	$0.74\scriptstyle{\pm 0.36}$
LLM only	$0.20\scriptstyle{\pm 0.13}$	$0.27\scriptstyle{\pm 0.16}$	$15.05\scriptstyle{\pm 3.49}$	$0.68\scriptstyle{\pm 0.19}$

Table 3: Action prediction performance comparison on the imputed dataset across different methods

Neuro-Symbolic Reasoning

Neuro-symbolic reasoning combines learned neural models with explicit logic, rules, knowledge structures, or executable tools to solve structured tasks.

See all 279 papers

DUPLEX: Agentic Dual-System Planning via LLM-Driven Information Extraction
Mar 25, 2026Keru Hua, Ding Wang, Yaoying Gu +1Robot Task Planning LLM Agent Planning
Midea Corporate Research Center, Midea Group, Foshan, China · Robot Science and Engineering Department, Northeastern University, Shenyang, China
When Planning Problems Become Hard for LLMs: Separating Structural Complexity from Semantic Interpretability
Oct 18, 2026Mayukh Debnath, Adeyinka Adedigba, Rammohan MallipeddiLLM Planning Symbolic Planning
Robotics and Autonomous Systems (AI), Arizona State University, Tempe, USA · Department of Artificial Intelligence, Kyungpook National University, Daegu, South Korea
From Solver Feedback to Faithful Plans: Multi-Role Reinforcement Learning for Symbolic Planning
Aug 22, 2026Chenghao Zhang, Yikai Mao, Haoyu Gao +4LLM Planning Symbolic Planning
University of Pennsylvania · Georgia Institute of Technology · Pace University +1

Mar 25, 2026J/K move · Enter open · S save

DUPLEX: Agentic Dual-System Planning via LLM-Driven Information Extraction

Keru Hua, Ding Wang, Yaoying Gu, Xiaoguang Ma

Midea Corporate Research Center, Midea Group, Foshan, China · Robot Science and Engineering Department, Northeastern University, Shenyang, China

Robot Task Planning LLM Agent Planning Neuro-Symbolic Planning Neuro-Symbolic Reasoning Symbolic Planning

Open paper AskPDF arXiv

While Large Language Models (LLMs) provide semantic flexibility for robotic task planning, their susceptibility to hallucination and logical inconsistency limits their reliability in long-horizon domains. To bridge the gap between unstructured environments and rigorous plan synthesis, we propose DUPLEX, an agentic dual-system neuro-symbolic architecture that strictly confines the LLM to schema-guided information extraction rather than end-to-end planning or code generation. In our framework, a feed-forward Fast System utilizes a lightweight LLM to extract entities, relations etc. from natural language, deterministically mapping them into a Planning Domain Definition Language (PDDL) problem file for a classical symbolic planner. To resolve complex or underspecified scenarios, a Slow System is activated exclusively upon planning failure, leveraging solver diagnostics to drive a high-capacity LLM in iterative reflection and repair. Extensive evaluations across 12 classical and household planning domains demonstrate that DUPLEX significantly outperforms existing end-to-end and hybrid LLM baselines in both success rate and reliability. These results confirm that The key is not to make the LLM plan better, but to restrict the LLM to the part it is good at - structured semantic grounding - and leave logical plan synthesis to a symbolic planner.

Abstract

Neuro-Symbolic Reasoning

Explore similar work

DUPLEX: Agentic Dual-System Planning via LLM-Driven Information Extraction

When Planning Problems Become Hard for LLMs: Separating Structural Complexity from Semantic Interpretability

From Solver Feedback to Faithful Plans: Multi-Role Reinforcement Learning for Symbolic Planning