Selected work

Representative Papers

MILP-SAT-GNN: Yet Another Neural SAT Solver

Jul 02, 2025

This work investigates the generalization capability and theoretical foundations of Graph Neural Networks (GNNs) for solving Boolean satisfiability (SAT) problems. Method: We model k-CNF formulas as weighted bipartite graphs and establish a reversible logical-to-graph mapping via mixed-integer linear programming encoding. To overcome expressivity limitations of standard GNNs on foldable formulas, we introduce Random Node Initialization (RNI); we further prove that, even without RNI, GNNs possess universal approximation power for unfoldable formulas. Contribution/Results: Theoretically, we establish clause-variable permutation invariance for the first time and characterize the completeness boundary of GNNs in SAT solving. Empirically, even simple GNN architectures achieve high accuracy and near-completeness in SAT classification under limited training data, demonstrating the feasibility and robustness of end-to-end neural SAT solving.

0 citationsRead paper

Novel Benchmark for NER in the Wastewater and Stormwater Domain

Jun 02, 2025

This work addresses the lack of benchmark datasets and effective methods for multilingual (French–Italian) named entity recognition (NER) in wastewater and stormwater management. We introduce the first bilingual domain-specific NER benchmark dataset for this field. To construct it, we propose an automated annotation protocol integrating cross-lingual label projection and domain terminology enhancement, combined with domain-adaptive pretraining, multilingual alignment, and zero-/few-shot fine-tuning of large language models (LLMs). Experiments show that cross-lingual projection achieves 89.2% labeling accuracy; LLM fine-tuning substantially outperforms traditional CRF and BiLSTM baselines. The publicly released corpus enables reproducible baseline evaluation. This study establishes a new benchmark and methodological foundation for low-resource, multilingual structured knowledge extraction and intelligent environmental governance.

0 citationsRead paper

Recent publications

Latest Papers

MILP-SAT-GNN: Yet Another Neural SAT Solver

Jul 02, 2025

This work investigates the generalization capability and theoretical foundations of Graph Neural Networks (GNNs) for solving Boolean satisfiability (SAT) problems. Method: We model k-CNF formulas as weighted bipartite graphs and establish a reversible logical-to-graph mapping via mixed-integer linear programming encoding. To overcome expressivity limitations of standard GNNs on foldable formulas, we introduce Random Node Initialization (RNI); we further prove that, even without RNI, GNNs possess universal approximation power for unfoldable formulas. Contribution/Results: Theoretically, we establish clause-variable permutation invariance for the first time and characterize the completeness boundary of GNNs in SAT solving. Empirically, even simple GNN architectures achieve high accuracy and near-completeness in SAT classification under limited training data, demonstrating the feasibility and robustness of end-to-end neural SAT solving.

0 citationsRead paper

Novel Benchmark for NER in the Wastewater and Stormwater Domain

Jun 02, 2025

This work addresses the lack of benchmark datasets and effective methods for multilingual (French–Italian) named entity recognition (NER) in wastewater and stormwater management. We introduce the first bilingual domain-specific NER benchmark dataset for this field. To construct it, we propose an automated annotation protocol integrating cross-lingual label projection and domain terminology enhancement, combined with domain-adaptive pretraining, multilingual alignment, and zero-/few-shot fine-tuning of large language models (LLMs). Experiments show that cross-lingual projection achieves 89.2% labeling accuracy; LLM fine-tuning substantially outperforms traditional CRF and BiLSTM baselines. The publicly released corpus enables reproducible baseline evaluation. This study establishes a new benchmark and methodological foundation for low-resource, multilingual structured knowledge extraction and intelligent environmental governance.

0 citationsRead paper