question decomposition

This work addresses the semantic ambiguity and missing constraint issues in Text-to-SQL systems when handling complex queries, which often stem from insufficient contextual understanding. To tackle these challenges, the paper proposes PV-SQL, a novel framework that integrates active database probing with rule-driven verification. The Probe component iteratively generates exploratory queries to dynamically retrieve database records, thereby clarifying value formats, column semantics, and inter-table relationships. Concurrently, the Verify component extracts verifiable conditions to construct an executable checklist, enabling iterative refinement of the generated SQL. Evaluated on the BIRD benchmark, PV-SQL improves execution accuracy by 5% and effective efficiency score by 20.8%, while substantially reducing the number of tokens required during inference.

This work addresses the challenge of integrating and retrieving multi-source heterogeneous data arising from schema inconsistencies by proposing an “executable schema contract” mechanism. This approach enables structure-aware automatic knowledge graph construction through a combination of closed-world field catalogs, deterministic structural analysis (e.g., primary/foreign key detection and source hierarchy identification), and monotonic extension protocols. It integrates large language model–constrained schema discovery, schema-guided information extraction and deduplication, and a multi-tool agent routing strategy that supports structured queries, graph traversal, and vector search. Evaluated on four question-answering benchmarks, the method achieves significantly superior zero-shot performance compared to pure retrieval and decomposition-based baselines. Ablation studies confirm that schema-conditioned routing, structural reasoning, and schema-guided construction are critical to its performance gains.

This work addresses the challenge of synthesizing modular specifications for client programs that use complex libraries—specifications that must simultaneously satisfy formal verification requirements and be amenable to empirical testing. To this end, the paper introduces a novel form of specification called *contextual contracts* and develops a counterexample-guided learning framework that integrates constraint solving with black-box testing. The approach leverages an ICE-learning-based generalization engine within a CHC (Constrained Horn Clauses) solver to co-synthesize concise, verifiable, and testable contracts for library methods along with inductive invariants. Experimental evaluation demonstrates that the implemented tool, vmtlc, successfully generates effective contracts meeting both verification and testing criteria on real-world client benchmarks.

This work addresses the absence of solver- and domain-agnostic verification mechanisms in Semantic-Guided Synthesis (SemGuS). Methodologically, it rigorously reduces correctness checking of SemGuS solutions to validity checking in Constraint Logic Programming (CLP), uniformly supporting first-order logic, constrained/coinstrained Horn clauses, and general CLP queries; it further extends the SemGuS syntax to accommodate nondeterministic and reactive synthesis. The key contributions are: (i) the first sound and complete reduction of SemGuS verification to CLP validity, thereby overcoming prior expressiveness limitations; (ii) enabling verification of previously inexpressible complex synthesis instances; and (iii) integration into an enumerative solver, achieving successful synthesis on benchmark problems unsolved by all existing SemGuS solvers. This framework establishes a foundational, general-purpose verification infrastructure for SemGuS.

This study addresses the high false positive and false negative rates in existing static analysis tools for detecting atomicity violations in smart contracts, which stem from insufficient context awareness and inadequate modeling of intermediate states. To overcome these limitations, the authors propose PSR², a novel framework that integrates control flow graphs and abstract syntax trees to construct a unified model of atomicity inconsistencies. PSR² employs a graph-structure analysis module to identify suspicious execution paths, a semantic context analysis module to extract data dependencies, and a fusion decision module that performs formal cross-validation. Evaluated on 1,600 real-world smart contracts, PSR² achieves an F1-score of 94.69%, substantially outperforming baseline tools (51.86%). Ablation studies further confirm that its fusion mechanism reduces false positives by nearly 50%.

First-order Answer Set Programming (ASP) lacks systematic support for modular and parameterized subroutines, hindering structured program design. This work proposes a formal framework for parameterized modular logic programs, introducing for the first time a parameterization mechanism and intensional declarations into modular ASP. By means of precise semantic mappings, it faithfully captures the collective control mechanisms employed in clingo. The approach establishes a theoretical foundation for modular ASP while enabling declarative definition, reuse, and clear semantics for modules. Consequently, it effectively bridges the gap between modularity and traditional ASP, demonstrating strong expressiveness and practical utility in program structuring and instantiation.

This work addresses the common challenge faced by data analysts who begin with vague questions and iteratively refine their information needs through exploration. To support this process, the authors propose Pneuma-Seeker, a novel system that leverages large language models (LLMs) not as opaque question-answering engines but as transparent, interactive analytical collaborators. Pneuma-Seeker enables users to explicitly articulate their information requirements as verifiable relational specifications and supports iterative refinement of these specifications, targeted data discovery, and execution with full provenance tracking. By integrating LLMs with relational specification modeling, data provenance, and an interactive interface, the system demonstrates its effectiveness in two real-world public procurement use cases, where it successfully helped users dynamically concretize evolving analytical needs and accurately retrieve relevant data.

This work addresses the high computational cost of large language models during test-time inference for formal theorem proving. The authors propose a verifier-guided learning refinement framework that, for the first time, leverages compact, structured failure patterns from compiler outputs to steer efficient tree search. By focusing on local error correction, the method avoids accumulating lengthy proof histories, thereby enabling more effective exploration. Integrating compiler feedback, tree search, and verifier signals, the approach significantly enhances reasoning efficiency. Evaluated on the PutnamBench benchmark, it achieves state-of-the-art results among publicly reported systems under comparable computational budgets, markedly improving baseline prover performance on both 8B and 32B parameter models.

Traditional data provenance struggles to capture the uncertainty arising from semantic ambiguities in multi-outcome systems, such as concurrent transactions or Datalog with negation. This work proposes a “deterministic provenance” framework that quantifies the dependency depth of query results on hierarchical semantic decisions by tracking the semantic commitments required to resolve ambiguities. Leveraging an exchange semiring structure over support sets and a filtration mechanism, the framework extends provenance to multi-outcome settings for the first time, unifying the modeling of provenance behavior across diverse semantic variants—including isolation levels and negation handling. It further enables composable robustness analysis of semantic decision paths and quantitative diagnosis of resolution costs.