Policy-Conditioned Constrained Decoding for Column-Level Access Control in Text-to-SQL

2026-07-14 · Ryoto Miyamoto, Xin Fan, Hayato Yamana arxiv

Text-to-SQL is increasingly deployed across trust boundaries between data providers and users. Such deployment must balance three competing requirements: policy compliance, answer coverage, and bounded cost. Existing approaches typically decide refusal based on which columns a query mentions and enforce it stochastically. Whether a query is compliant, however, depends not only on which columns appear but on how they are used, and stochastic enforcement cannot deterministically rule out violations. We formalize this requirement as a column-use policy over semantic use: output, filter condition, and aggregation argument. We integrate the policy by aligning each role with grammar productions tracked by the decoder. The resulting system, PCC-SQL, applies a per-token logits mask that deterministically eliminates single-query column-use violations on the supported SQL fragment in a single decoding pass. Across three benchmarks and three open-source models, PCC-SQL achieves 0% Leakage Rate and Coverage up to 88.7% on Spider-CU, while staying within +10% tokens of direct prompting. We additionally assess semantic alignment with execution accuracy.

📄 PDF Abstract BibTeX arXiv:2607.12341

Code (1)

arxivsub/arXivSub_daily_arxiv ★ 3

Results from the Paper

Rank	Task	Dataset	Model	Metrics
#7	Text-To-SQL	spider	Policy-Conditioned Constrained Decoding	Execution Accuracy (Test): 88.7

Learning Adaptive LLM Decoding

2026-03-10 · Chloe H. Su, Zhe Ye, Samuel Tenka, Aidan Yang 외 arxiv

Decoding from large language models (LLMs) typically relies on fixed sampling hyperparameters (e.g., temperature, top-p), despite substantial variation in task difficulty and uncertainty across prompts and individual dec…

Reinforcement Learning

The Hidden Cost of Structured Generation in LLMs: Draft-Conditioned Constrained Decoding

2026-02-08 · Avinash Reddy, Thayne T. Walker, James S. Ide, Amrit Singh Bedi arxiv

Large language models (LLMs) are increasingly used to generate executable outputs, JSON objects, and API calls, where a single syntax error can make the output unusable. Constrained decoding enforces validity token-by-to…

Goal-Conditioned Reinforcement Learning with Imagined Subgoals

2021-07-01 · Elliot Chane-Sane, Cordelia Schmid, Ivan Laptev

Goal-conditioned reinforcement learning endows an agent with a large variety of skills, but it often struggles to solve tasks that require more temporally extended reasoning. In this work, we propose to incorporate imagi…

reinforcement-learning Reinforcement Learning Reinforcement Learning (RL)

Constrained Prescriptive Trees via Column Generation

2022-07-20 · Shivaram Subramanian, Wei Sun, Youssef Drissi, Markus Ettl

With the abundance of available data, many enterprises seek to implement data-driven prescriptive analytics to help them make informed decisions. These prescriptive policies need to satisfy operational constraints, and p…

Guided Cooperation in Hierarchical Reinforcement Learning via Model-based Rollout

2023-09-24 · Haoran Wang, Zeshen Tang, Leya Yang, Yaoru Sun 외

Goal-conditioned hierarchical reinforcement learning (HRL) presents a promising approach for enabling effective exploration in complex, long-horizon reinforcement learning (RL) tasks through temporal abstraction. Empiric…

Hierarchical Reinforcement Learning reinforcement-learning Reinforcement Learning (RL)