VLP lifts LLM code pass rates from 29-73% to 65-93%
Guiding Human Validation of LLM-Generated Code via Verifiable Literate Programming
Unambiguous natural-language docs let users spot and fix intent mismatches before verification
full image
Programming Languages
Covers programming language semantics, language features, programming approaches (such as object-oriented programming, functional programming, logic programming). Also includes material on compilers oriented towards programming languages; other material on compilers may be more appropriate in Architecture (AR). Roughly includes material in ACM Subject Classes D.1 and D.3.
Guiding Human Validation of LLM-Generated Code via Verifiable Literate Programming
Unambiguous natural-language docs let users spot and fix intent mismatches before verification
full image
Kani: A Model Checker for Rust
Translates MIR to CBMC to prove functional correctness and absence of panics beyond what the type system guarantees.
full image
Trustworthy Runtime Verification via Bisimulation (Extended Experience Report)
Generated proofs confirm matching outputs and crash behavior for each monitor using symbolic execution of LLVM.
full image
Practical Range Refinement Types with Inference
Ranger verifies index safety in loops and arrays using bidirectional typing plus flow analysis and needs fewer annotations than prior tools.
full image
LLVM-Bench: Benchmarking and Advancing Large Language Models for LLVM Compiler Issue Resolution
New benchmark of 423 real tasks shows single models lag on patches and builds but combining outputs raises success
full image
Certificate-Carrying Transformation of Event-Driven Block Programs
For Scratch programs, recomputing every side condition ensures behavior preservation even if the optimizer is buggy.
full image
Fixed-Set Robustness in Programming by Example: Example Corruption and Semantic Partition Recovery
Semantic partition recovery works only when clean semantics already carry a decisive vote margin, which is rare on standard benchmarks.
full image
AlgoBench: Benchmarking Algorithmic Adaptation in Code Generation
AlgoBench transforms competitive programming tasks so original solutions fail, exposing limits in algorithmic adaptation.
full image
A Multi-Dimensional, Per-Pass Empirical Study of the LLVM Optimization Pipeline
Per-prefix measurements on 30 kernels show most gains arrive late and the final config loses on size-speedup for 29 kernels.
full image
The Fourth-Root Complexity of Data Movement
Abstract memory hierarchy shows per-access costs scale as N to the 1/4 for common apps, distinguishing power-law from exponential miss ratio
Binary logical relations prove preservation once let-insertion is tracked as a control effect in two-stage calculi.
full image
NSynC: Normalised Synthesis of Computation
Enumerating semantics directly avoids evaluating duplicate programs that differ only in syntax.
full image
By representing data dependencies explicitly, proposals update only the relevant subgraph, increasing samples per unit time.
full image
Langshaw: Declarative Interaction Protocols Based on Sayso and Conflict
Sayso assigns attribute control while nono and nogo handle action conflicts for flexible multiagent enactment.
full image
Symbolon: Symbolic Execution by Learning Code Transformation
Distilling search results from small programs into agent skills cuts memory and solver time while exposing new kernel bugs.
full image
KoAT: Automatic Complexity and Termination Analysis of Integer Programs
Modular analysis of subprograms with multiple techniques yields termination proofs and complexity results for recursive code.
full image
Prophecy-Based Automated Verification of Message-Passing Programs
Future-value lists plus timestamps let sender and receiver threads be encoded separately while preserving exact correctness.
full image
Same Coeffect, Different Base: Connecting Two Dominant Approaches to Graded Types
Graded-base and linear-base systems model the same context dependencies via type-grade-semantics preserving mappings.
full image
It models computation-quantum interactions via a new integral on type I algebras and is shown to be a strong monad in both finitary and meas
The principle of tying functionality to domain objects leads to inevitable problems that later methods cannot fix.
full image
Welterweight Go: Boxing, Structural Subtyping, and Generics (Extended Version)
Type-directed compilation to a runtime model preserves separate compilation and avoids code generation.
full image
Mostly Automatic Translation of Language Interpreters from C to Safe Rust
Six programs pass all original tests and most validation tests while losing their memory bugs.
full image
Formalizing a Many-Sorted Hybrid Polyadic Modal Logic in Lean
The formalization includes a soundness proof and a DSL to define and check programs and security protocols without extra syntax.
Compiler-Driven Approximation Tuning for Hyperdimensional Computing
ApproxHDC searches the space of possible approximations to deliver performance gains on CPUs, GPUs, and memory accelerators with little accu
full image
Theory-Scale Auto-Formalization of Logics for Computer Science
Dataset of 327 interdependent items and 85K lines of Lean code tests coherent translation of definitions and theorems.
full image
An Empirical Study of LLM-Generated Specifications for VeriFast
Empirical test on 303 C functions shows errors mostly come from missing separation logic knowledge.
full image
A Typestate Approach to Purpose-aware Programming
PurPL represents allowed uses as type states that grow or shrink, enabling static checks on sensitive data handling in object-oriented code.
full image
Axon: A Synthesizing Superoptimizer for Tensor Programs
Axon synthesizes kernels for AI accelerators by propagating operators and checking equivalence over unbounded tensor domains.
full image
A Survey of Quantum Programming Languages
By comparing ten popular languages, the survey pinpoints issues future quantum programming tools must solve to match classical ones.
full image
Information flow security on persistent memory
The condition used for x86 out-of-order execution extends to crash-resilient memory propagation rules.
Reading AI Model Compilation in MLIR Through the Lens of Formal Theories
Formal correspondences clarify completeness and trade-offs as coding agents handle more implementation.
full image
Verifiable Auto-Formalization of Mathematics Using a Relaxed Natural Formal Language
It keeps human reasoning structure and defers precise interpretation to a constrained elaboration step followed by tactic scripts.
full image
DissProve: Automated Verification of Distributed Protocols with Affine Communication
Affine protocols admit goal-directed backward proofs that avoid global invariants and handle any number of participants.
full image
ESBMC-PLC+: A Unified IEC 61131-3 Formal Verification Framework as a PLCverif Successor
ESBMC-PLC+ adds Ladder Diagram support and k-induction to match PLCverif coverage while proving safety for any cycle count.
full image
A Compositional Language for Property Graphs
Cleaned path queries with variables and #Datalog allow full mixing of graph and relational queries while building new graphs.
full image
Cyclic Graphs and Memoization in Pure λ-Calculus
Weak-head tabling produces finite sound graphs for any finite-state term and performs automatic memoization.
full image
Lifting E-Graphs: A Function Isn't a Constant
Thinning bitvectors and a modified union-find keep alpha-canonical variables rigid inside the structure.
full image
Compositional Generator Equivalence (Extended Version)
Arrow-calculus restriction regains a usable compositional distribution semantics while capturing practical generators.
full image
Powerdomains and nondeterminism in synthetic domain theory
The constructions give computationally adequate models that embed into dependent type theory for reasoning about programs.
full image
Analyzing the Analyzers: Model Counting Meets Abstract Interpretation
Encodes concrete and abstract semantics as formulas to compute semantic loss directly, revealing intervals rival octagons and bit-level doma
full image
Shared-Context Batched Satisfiability
Over-approximation wins easy symbolic abstraction cases while CLF handles more hard property checks.
full image
CNnotator: LLM-Guided Memory Safety Annotation Synthesis
CNnotator automates annotation of implicit memory usage, reaching high success rates on small-to-medium code and suggesting broader practica
full image
Correct and Complete Symbolic Execution for Free
One rule format produces matching concrete and symbolic semantics without separate development.
full image
The Alignment Problem in Constrained Code Generation
Unconstrained decoding reaches higher functional correctness by avoiding low-probability regions created when constrainers reject valid prog
full image
KBSpec: LLM-driven Formal Specification Generation with Evolving Domain Knowledge Base
A knowledge base updated from verifier feedback on successful JML generations improves completeness without model tuning.
full image
Compiling Differentiable Audio Graphs to Real-Time DSP
The generated real-time code replicates the model's impulse response within floating-point noise and includes built-in stability checks.
full image
AI-Assisted Completion of CertiGC Proofs: An Experience Report
Recorded-backward-edge requirement replaces invalidated no-backward-edge assumption and restores the top-level correctness theorem.
full image
Big-step and small-step Horn clause derivations applied to operational semantics
Interpreter specialization yields equivalent clause sets, turning big-step language semantics into small-step versions.
Multi-LCB: Extending LiveCodeBench to Multiple Programming Languages
Extending LiveCodeBench to twelve languages uncovers performance gaps while retaining contamination controls.
full image
Generalized interchange axiom for finite distributions enables proofs of semi-graphoid axioms, do-calculus and Markov category properties wi
An MSO Framework for Weak-Memory Verification and Robustness
This split lets MSO express some memory models for uniform verification and robustness checks but rules out others unless a quadratic-time p
Effect Systems as Abstract Interpretations
Embedding effect quantales shows they arise from interpreting event sequences rather than program states.
full image
EEG case study shows consistent interpretable traces from simulator to IBM hardware without circuit authoring.
full image
OpenRath: Session-Centered Runtime State for Agent Systems
Making fork, merge, and replay explicit operations instead of external trace reconstructions.
full image
Graphical exports from CONTROLLINO and OpenPLC Editor now produce full GOTO programs that the SMT checker can examine instead of empty IR.
full image
VERITAS: Verifier-Guided Proof Search for Zero-Shot Formal Theorem Proving
Two-phase protocol turns Phase-1 failures into MCTS negatives and adds solves on both miniF2F and a new combinatorics set without losing ear
full image
Behind Python: The Languages That Power AI
Five algorithms benchmarked in six languages show Rust close behind and rankings that shift by workload.
full image
Defense-in-Depth Runtime Safety in Move
Aptos blockchain performs independent execution-time verification of invariants to block asset loss from static analysis flaws.
full image
Bridging Functional Correctness and Runtime Efficiency Gaps in LLM-Based Code Translation
Two stages generate diverse candidates then select the optimal by comparing differences, with gains on three benchmarks.
full image
Visored: A Controlled-Natural-Language Prover for LLM-Generated Mathematics
A rule-driven automation layer fills omitted steps so accepted proofs export as verified Lean code, and LLMs use the system on miniF2F witho
full image
When the Next Step Is Not One Step: Distribution-Aware Execution Modeling for Concurrent Go Programs
Fewer than 1000 traces let a 7B model beat zero-shot Gemini while cutting calibration error on real concurrency bugs
full image
Verifying the Rust Standard Library
An open campaign folds multiple tools into CI to produce machine-checked guarantees against out-of-bounds access and related undefined behav
full image
A Stone-Cech Collecting Semantics for Residual Process Behaviour
Tail-cluster sets of residual streams distinguish stable, recurrent, mixed and escaping nontermination while validating prefix and choice la
A verified chain L0 to L4 supplies sound prevention for stale-generation, phantom-tool, causal-cascade and tool-effect reordering.
full image
PromptMN: Pseudo Prompting Language
PromptMN annotates natural language with typed instructions so models resolve roles, goals, conditionals, and prime checks reliably.
full image
ScratchLens: Lens-Parametric Behavioral Equivalence for Scratch Programs
Causal IR plus trace quotients classify all 444 validated pairs from real projects and record none on 158 witnessed-different ones.
full image
Scalable Probabilistic Program Verification via Typed Extended Decision Diagrams
TEDDs keep weakest pre-expectations compact on loops, enabling SMT pruning where prior methods ran out of memory.
full image
ANEForge: Python for direct computation on the Apple Neural Engine
Compiles lazy tensor graphs to native ANE programs that finish ResNet-18 forward passes in 0.33 ms and support training.
full image
Specifying Hardware Communication as Programs
A DSL encodes a protocol once as an imperative script; the same script generates signals and reconstructs transactions from waveforms.
full image
Probabilistic, Resource-Aware, Asynchronous, Out-of-Order Choreographies
Programs yield networks that track values and availability times under delays, failures, and recovery.
full image
A Dual-Mode Faust-to-CLAP Compilation System
Static and dynamic modes keep parameters stable during code changes without host restarts.
full image
Functional Cache Grafting: Robust and Rapid Code-Policy Synthesis for Embodied Agents
FCGraft retrieves and stitches validated code segments with their caches to build reliable control programs more efficiently than full-promp
full image
nomp: A Framework for Building Domain Specific Compilers
Pragma model plus runtime aims to reuse proven patterns so productivity rises without losing performance or portability
full image
Web-Native Graphical EMF Model Editors
The Angular apps handle references and containment exactly like desktop EMF tools but require no server.
full image
Defeat the Heap: Zero-Copy Data Movement in AXI4MLIR
AXI4MLIR now places buffers directly in DMA-mapped regions and skips the heap staging copy.
full image
Dynamic Software Updates using CRDTs
Compatible middleware lets distributed components at mixed versions keep communicating during backwards-incompatible changes.
full image
Max-Policy Iteration, Revisited
For integers and floats, it still terminates and enables precise bounds without widening.
full image
Answer Set Programming for Egg Extraction and More
Naive approach matches optimized ILP and recovers extra optima on complex instances
full image
Modeling Components and Connections in Cyber-Physical Systems
Developers draw nodes, publishers and links with automatic validation instead of writing text configurations.
When Types Intersect and Effects Get Handled
The system also converts reachability into type inference and produces a decidable simple type system.
full image