paper-with-me

Papers

PAC Learning-Based Verification and Model Synthesis

2015-11-03 · Yu-Fang Chen, Chiao Hsieh, Ondřej Lengál, Tsung-Ju Lii, Ming-Hsien Tsai, Bow-Yaw Wang, Farn Wang

We introduce a novel technique for verification and model synthesis of sequential programs. Our technique is based on learning a regular model of the set of feasible paths in a program, and testing whether this model contains an incorrect behavior. Exact learning algorithms require checking equivalence between the model and the program, which is a difficult problem, in general undecidable. Our learning procedure is therefore based on the framework of probably approximately correct (PAC) learning, which uses sampling instead and provides correctness guarantees expressed using the terms error probability and confidence. Besides the verification result, our procedure also outputs the model with the said correctness guarantees. Obtained preliminary experiments show encouraging results, in some cases even outperforming mature software verifiers.

📄 PDF Abstract BibTeX arXiv:1511.00754

Code (0)

등록된 구현이 없습니다.

Tasks

modelPAC learning

Similar Papers 제목 키워드 기반

Invariant Synthesis for Incomplete Verification Engines

2017-12-15 · Daniel Neider, Pranav Garg, P. Madhusudan, Shambwaditya Saha 외

We propose a framework for synthesizing inductive invariants for incomplete verification engines, which soundly reduce logical problems in undecidable theories to decidable theories. Our framework is based on the counter…

Marco DeepResearch: Unlocking Efficient Deep Research Agents via Verification-Centric Design

2026-03-30 · Bin Zhu, Qianghuai Jia, Tian Lan, Junyang Ren 외 arxiv

Deep research agents autonomously conduct open-ended investigations, integrating complex information retrieval with multi-step reasoning across diverse sources to solve real-world problems. To sustain this capability on …

Information Retrieval

Toward Neural-Network-Guided Program Synthesis and Verification

2021-03-17 · Naoki Kobayashi, Taro Sekiyama, Issei Sato, Hiroshi Unno

We propose a novel framework of program and invariant synthesis called neural network-guided synthesis. We first show that, by suitably designing and training neural networks, we can extract logical formulas over integer…

Program Synthesis

What Do Claim Verification Datasets Actually Test? A Reasoning Trace Analysis

2026-04-02 · Delip Rao, Chris Callison-Burch arxiv

Despite rapid progress in claim verification, we lack a systematic understanding of what reasoning these benchmarks actually exercise. We generate structured reasoning traces for 24K claim-verification examples across 9 …

Arithmetic Reasoning

Polarimetric Thermal to Visible Face Verification via Self-Attention Guided Synthesis

2019-04-15 · Xing Di, Benjamin S. Riggan, Shuowen Hu, Nathaniel J. Short 외

Polarimetric thermal to visible face verification entails matching two images that contain significant domain differences. Several recent approaches have attempted to synthesize visible faces from thermal images for cros…

Face VerificationGenerative Adversarial NetworkImage Generation