Compiler-Guided Program Synthesis

Follow this topic Ask about this topic

Compiler-guided program synthesis studies how to turn specifications, examples, or observed behavior into executable programs using compilers, type checkers, formal verifiers, and tests as feedback. Common approaches combine constrained generation with counterexample-guided repair, bounded search, and language-model agents that use diagnostics or recompilation; explicit resource models and proof obligations can add stronger guarantees. Applications include API-use code, planning heuristics, SQL and analytical workflows, verified systems, and recovering source code from binaries, evaluated through compilation, execution, correctness, proofs, and benchmark task success. A central tension is balancing guarantees and generalization against search and verification costs: passing compiler checks alone may not establish intended semantics, and feedback can be incomplete or too specific.

Latest papers

16 papers5 papers in the last four weeks, against 2 the four weeks before. 0.0% of all new papers.

Compile Away Uncertainty: Self-Evolving Agent Skills via Programmatic Compilation
Oct 29, 2026Chengyu Yang, Yunxiang Zheng, Zongzheng Li +6LLM Agent Skill Learning Self-Evolving Agents
East China Normal University · China University of Geoscience
SCOPE: Compiling Natural-Language Policies into Neurosymbolic Predicate Programs
Oct 22, 2026Arnon Mazza, Elad LeviLLM Safety Monitoring AI Agent Governance
Plurai
Learning to Search for Executable Patternmaking from Certified Programs
Oct 15, 2026Zhuotao Xu, Chi Zeng, Liying Yang +2Inference-Time Search Compiler-Guided Program Synthesis
Macau University of Science and Technology
LabCompiler: A Framework for Verifiable Human-Machine Laboratory Execution from Imperfect Protocols
Oct 14, 2026Jingyuan Li, Dongzhan Zhou, Minting Pan +10Laboratory Automation Human-in-the-Loop AI
Shanghai Jiao Tong University · Shanghai Artificial Intelligence Laboratory · Fudan University +2
Compiling Expert Procedures into Verified Programs: When a Small Model Is Enough
Oct 13, 2026Vishvesh Bhat, Asad Aali, Muhammad Ahmed MohsinLLM Tool Use LLM-Based Program Synthesis
CoreThink AI · Stanford University
ProCARE: Real-World Study Automation via Profile-Grounded Evidence Contracts
Oct 12, 2026Zhaoyu Fan, Bowen Han, Jincheng Ou +8Scientific Workflow Automation Evidence-Grounded Generation
Tongji University · Shanghai Hengfang Health Technology Co., Ltd. · University of Electronic Science and Technology of China +1
Djinnlang: Higher-Level Programming by Unambiguous Specification with an LLM in the Compiler
Sep 21, 2026Simon Henniger, Stephen Chong, Nada AminLLM-Based Program Synthesis Compiler-Guided Program Synthesis
Harvard University Cambridge, Massachusetts, USA
Echo: Learning-based Matching Decompilation using Trusted Back Translation
Sep 16, 2026Jun Bi, Xiangxin Fang, Aarsh Chaube +3Software Reverse Engineering Program Synthesis
State Key Lab of Processors, Institute of Computing Technology, Chinese Academy of Sciences · University of Edinburgh
Bounded Semantic Planning and Deterministic Compilation for Reliable Enterprise Text-to-SQL
Aug 17, 2026Yi AiText-to-SQL Semantic Parsing
Agentic Synthesis against Counterexample-Supplemented Sketches
Jul 17, 2026Muness Castle, Eric RubeckCoding Agents LLM-Based Program Synthesis
Independent
Knowledge-Centric Agents for Workflow Generation in ComfyUI
Jul 17, 2026Zhendong Li, Lei Sun, Ruibo Ming +4Agentic Workflows LLM-Based Program Synthesis
INSAIT, Sofia University “St. Kliment Ohridski” · Adobe Research
Minimalist Genetic Programming
Jun 8, 2026Leonardo TrujilloGenetic Programming Program Synthesis
Tecnol´ogico Nacional de M´exico/IT de Tijuana,Tijuana, BC, Me xico · LASIGE, Department of Informatics, Faculty of Sciences, University of Lisbon, Lisbon, Portugal
CP-Agent: A Calibrated Risk-Controlled Agent for Feedback-Driven Competitive Programming
May 23, 2026Peisong Wang, Bowen Liu, Zehua Li +4Code Generation LLM-Based Program Synthesis
The Hong Kong University of Science and Technology (Guangzhou) Guangzhou, China
Inductive Deductive Synthesis: Enabling AI to Generate Formally Verified Systems
May 22, 2026Shubham Agarwal, Alexander Krentsel, Shu Liu +10AI Coding Agents Program Synthesis
1UC Berkeley · 2Google · 3UC Santa Cruz
Property-Guided LLM Program Synthesis for Planning
May 15, 2026André G. Pereira, Augusto B. Corrêa, Jendrik SeippSymbolic Planning Counterexample-Guided Refinement
Federal University of Rio Grande do Sul · Brazil · University of Oxford +3
Decaf: Improving Neural Decompilation with Automatic Feedback and Search
May 12, 2026Alexander Shypula, Osbert Bastani, Edward SchwartzSoftware Reverse Engineering Compiler-Guided Program Synthesis
University of Pennsylvania Philadelphia, Pennsylvania 19104 · Carnegie Mellon University Pittsburgh, Pennsylvania 15213
A Synthesis Method of Safe Rust Code Based on Pushdown Colored Petri Nets
Apr 2, 2026Kaiwen Zhang, Guanjun LiuProgram Synthesis Compiler-Guided Program Synthesis
Tongji University, Shanghai, China

Oct 29, 2026J/K move · Enter open · S save

Compile Away Uncertainty: Self-Evolving Agent Skills via Programmatic Compilation

Chengyu Yang, Yunxiang Zheng, Zongzheng Li, Jiamei Li+5

East China Normal University · China University of Geoscience

LLM Agent Skill Learning Self-Evolving Agents LLM-Based Program Synthesis Compiler-Guided Program Synthesis

Open paper AskPDF arXiv

Self-evolving agent skills enable AI agents to accumulate and refine reusable procedural knowledge from interaction experience without updating model parameters. However, since such skills are typically represented as textual documents and executed through LLM reasoning, even previously learned procedures must be repeatedly interpreted and regenerated, introducing avoidable uncertainty into every execution. In contrast, program execution is deterministic and independently verifiable. Intuitively, transforming well-specified operations into executable programs can reduce reliance on LLM reasoning and, consequently, reduce execution uncertainty. Motivated by this intuition, we introduce SKillCompiler, a self-evolving skill framework that integrates programmatic compilation into the skill evolution process. Specifically, SkillCompiler adopts an executability-aware skill compilation strategy that decomposes skills into fine-grained operations, analyzes their executability, and constructs hybrid skill skeletons that retain LLM reasoning for context-dependent operations while compiling well-specified ones into validated programs. To avoid redundant program synthesis and enable executable knowledge to accumulate across evolving skills, we further design a program asset reuse mechanism that organizes validated programs as reusable assets, retrieves existing implementations whenever applicable, and synthesizes and verifies new programs only when necessary. Extensive experiments on five benchmarks and six model families demonstrate that SkillCompiler consistently outperforms state-of-the-art methods, achieving up to 32.33% improvement in task performance.