Institution profile

MongoDB

Industry researchnorthamerica · us
Official website
Research library9linked papers
Opportunities26open roles
Selected work

Representative Papers

AptMQL-Bench: From Text-to-SQL to Text-to-MQL via Access-Pattern Schema Design and Data-Preserving Migration

Oct 01, 2026

This study addresses the limitations of existing text-to-MQL benchmarks, where mechanical translation causes information loss and query inefficiency, hindering natural language querying over document databases. We propose an access-pattern-driven construction paradigm that abandons heuristic conversion in favor of a human-in-the-loop workflow with coding agents. By designing native MongoDB schemas according to anticipated access patterns and performing query rewriting, our approach achieves lossless migration from SQLite to document databases. Based on this methodology, we introduce AptMQL-Bench, comprising 21 databases and 3,186 samples. Evaluations reveal that even the strongest model attains only 57.38% accuracy, demonstrating the substantial challenge and research value of this task.

0 citationsRead paper

Plaintext Recovery Against Post-Filtering Access Control

Aug 12, 2026

While existing post-filtering–based fine-grained access control (FGAC) mechanisms prevent direct data leakage, their susceptibility to existence side channels poses significant risks under complex queries. This work presents the first systematic demonstration of how timing and scoring side channels, combined with rich query interfaces—such as range, prefix, and conjunctive predicates—can be exploited to efficiently reconstruct high-entropy plaintext content in PostgreSQL and Elasticsearch/OpenSearch. By leveraging binary search, SQL expressions, prefix expansion, and document scoring mechanisms, we devise novel reconstruction attacks tailored to relational databases and search engines, successfully recovering unknown attribute values, complete records, and corpus n-grams. Our findings expose fundamental security flaws in current post-filtering FGAC schemes when deployed in environments supporting expressive query predicates.

0 citationsRead paper

Beyond Text Editing: Algebraic Manipulation of Source Code

Jul 21, 2026

Traditional text-based code editing struggles to satisfy syntactic and semantic constraints, leading to inefficient performance by large language model (LLM) agents when performing cross-file or non-local modifications. This work proposes Source Code Algebra (SCA), a novel framework that introduces compositional, idempotent, and commutative algebraic operations into code editing, replacing character-level edits with semantically complete atomic operations. Built upon abstract syntax trees, SCA defines semantics-aware edit primitives and formally models their algebraic properties, integrating them into an LLM interface. Experimental evaluation of the prototype system SCAS demonstrates that this paradigm significantly improves modification success rates while reducing communication token overhead by one to two orders of magnitude, highlighting its potential as an efficient and accurate new approach to code editing.

0 citationsRead paper

Autoregressive Synthesis of Sparse and Semi-Structured Mixed-Type Data

Mar 01, 2026

This work addresses the limitations of existing synthetic data generation methods, which rely on dense tabular assumptions and struggle to effectively model sparse, semi-structured, and mixed-type data such as JSON, thereby limiting their scalability in modern data systems. To overcome this, the authors propose Origami, an end-to-end architecture based on autoregressive Transformers that employs a structure-aware tokenization strategy. This approach natively encodes nested objects, variable-length arrays, and heterogeneous data types into sequences of key, value, and structural tokens—without requiring flattening or imputation. Experimental results demonstrate that Origami significantly outperforms baseline methods—including GANs, VAEs, and diffusion models—in terms of fidelity, utility, and privacy preservation, maintaining high-quality synthesis even on semi-structured data with sparsity levels as high as 38%.

0 citationsRead paper
Recent publications

Latest Papers

AptMQL-Bench: From Text-to-SQL to Text-to-MQL via Access-Pattern Schema Design and Data-Preserving Migration

Oct 01, 2026

This study addresses the limitations of existing text-to-MQL benchmarks, where mechanical translation causes information loss and query inefficiency, hindering natural language querying over document databases. We propose an access-pattern-driven construction paradigm that abandons heuristic conversion in favor of a human-in-the-loop workflow with coding agents. By designing native MongoDB schemas according to anticipated access patterns and performing query rewriting, our approach achieves lossless migration from SQLite to document databases. Based on this methodology, we introduce AptMQL-Bench, comprising 21 databases and 3,186 samples. Evaluations reveal that even the strongest model attains only 57.38% accuracy, demonstrating the substantial challenge and research value of this task.

0 citationsRead paper

Plaintext Recovery Against Post-Filtering Access Control

Aug 12, 2026

While existing post-filtering–based fine-grained access control (FGAC) mechanisms prevent direct data leakage, their susceptibility to existence side channels poses significant risks under complex queries. This work presents the first systematic demonstration of how timing and scoring side channels, combined with rich query interfaces—such as range, prefix, and conjunctive predicates—can be exploited to efficiently reconstruct high-entropy plaintext content in PostgreSQL and Elasticsearch/OpenSearch. By leveraging binary search, SQL expressions, prefix expansion, and document scoring mechanisms, we devise novel reconstruction attacks tailored to relational databases and search engines, successfully recovering unknown attribute values, complete records, and corpus n-grams. Our findings expose fundamental security flaws in current post-filtering FGAC schemes when deployed in environments supporting expressive query predicates.

0 citationsRead paper

Beyond Text Editing: Algebraic Manipulation of Source Code

Jul 21, 2026

Traditional text-based code editing struggles to satisfy syntactic and semantic constraints, leading to inefficient performance by large language model (LLM) agents when performing cross-file or non-local modifications. This work proposes Source Code Algebra (SCA), a novel framework that introduces compositional, idempotent, and commutative algebraic operations into code editing, replacing character-level edits with semantically complete atomic operations. Built upon abstract syntax trees, SCA defines semantics-aware edit primitives and formally models their algebraic properties, integrating them into an LLM interface. Experimental evaluation of the prototype system SCAS demonstrates that this paradigm significantly improves modification success rates while reducing communication token overhead by one to two orders of magnitude, highlighting its potential as an efficient and accurate new approach to code editing.

0 citationsRead paper

Autoregressive Synthesis of Sparse and Semi-Structured Mixed-Type Data

Mar 01, 2026

This work addresses the limitations of existing synthetic data generation methods, which rely on dense tabular assumptions and struggle to effectively model sparse, semi-structured, and mixed-type data such as JSON, thereby limiting their scalability in modern data systems. To overcome this, the authors propose Origami, an end-to-end architecture based on autoregressive Transformers that employs a structure-aware tokenization strategy. This approach natively encodes nested objects, variable-length arrays, and heterogeneous data types into sequences of key, value, and structural tokens—without requiring flattening or imputation. Experimental results demonstrate that Origami significantly outperforms baseline methods—including GANs, VAEs, and diffusion models—in terms of fidelity, utility, and privacy preservation, maintaining high-quality synthesis even on semi-structured data with sparsity levels as high as 38%.

0 citationsRead paper