https://arxiv.org/api/ddUEplPmVWhG4hlJj2pLO+khGn4 2026-06-14T04:14:30Z 9885 240 15 http://arxiv.org/abs/2511.10343v2 2026-05-01T10:33:38Z

The Damas-Hindley-Milner (ML) type system owes its success to principality, the property that every well-typed expression has a unique most general type. This makes inference predictable and efficient. Unfortunately, many extensions of ML (GADTs, higher-rank polymorphism, and static overloading) endanger princpality by introducing _fragile_ constructs that resist principal inference. Existing approaches recover principality through directional inference algorithms, which propagate _known_ type information in a fixed (or static) order (e.g. as in bidirectional typing) to disambiguate such constructs. However, the rigidity of a static inference order often causes otherwise well-typed programs to be rejected. We propose _omnidirectional_ type inference, where type information flows in a dynamic order. Typing constraints may be solved in any order, suspending when progress requires known type information and resuming once it becomes available, using _suspended match constraints_. This approach is straightforward for simply typed systems, but extending it to ML is challenging due to let-generalization. Existing ML inference algorithms type let-bindings (let x = e1 in e2) in a fixed order: type e1, generalize its type, and then type e2. To overcome this, we introduce _incremental instantiation_, allowing partially solved type schemes containing suspended constraints to be instantiated, with a mechanism to incrementally update instances as the scheme is refined. Omnidirectionality provides a general framework for restoring principality in the presence of fragile features. We demonstrate its versatility on two fundamentally different features of OCaml: static overloading of record labels and datatype constructors and semi-explicit first-class polymorphism. In both cases, we obtain a principal type inference algorithm that is more expressive than OCaml's current typechecker.

2025-11-13T14:15:36Z updated and improved version: 44 pages + appendices Alistair O'Brien Didier Rémy Gabriel Scherer http://arxiv.org/abs/2604.26967v2 2026-05-01T07:24:12Z

\emph{Literate programming}, introduced by Knurth, interleaves code and prose so that a program can be read as both executable and explanatory text. We propose \emph{literate execution}, which inverts this relationship: rather than embedding code within a static narrative, we treat documentation -- and other expository elements such as visualisations -- as first-class artefacts that can be computed alongside a running program and then integrated into a view of its execution. We explore this idea through Fluid, a programming language with a provenance-tracking runtime that records fine-grained dependencies between inputs and outputs. These provenance relationships can be surfaced as interactions that allow readers to explore how intermediate values contribute to a result. By integrating visualisation, provenance, and exposition, literate execution aims to make programs more explorable and self-explanatory, and explorable explanations easier to program.

2026-04-17T06:55:44Z A version of this paper was accepted to the 16th Annual Workshop on the Intersection of HCI and PL (PLATEAU 2026). https://2026.plateau-workshop.org/ Joe Bond Jacob Pake Cristina David Andrew McNutt Trevor Sseguya Muwonge Dominic Orchard Roly Perera http://arxiv.org/abs/2605.00314v1 2026-05-01T00:48:47Z

An agent skill is a configuration package that equips an LLM-driven agent with a concrete capability, such as reading email, executing shell commands, or signing blockchain transactions. Each skill is a hybrid artifact-a structured half declares executable interfaces, while a prose half dictates when and how those interfaces fire-and the prose is reinterpreted probabilistically on every invocation. Conventional static analyzers parse the structured half but ignore the prose; LLM-based tools read the prose but cannot reproducibly prove that a tainted input reaches a high-impact sink. We present Semia, a static auditor for agent skills. Semia lifts each skill into the Skill Description Language (SDL), a Datalog fact base that captures LLM-triggered actions, prose-defined conditions, and human-in-the-loop checkpoints. Synthesizing a fact base that is both structurally sound and semantically faithful to the original prose is the central challenge; we address it with Constraint-Guided Representation Synthesis (CGRS), a propose-verify-evaluate loop that refines LLM candidates until convergence. Security properties (e.g., indirect injection, secret leakage, confused deputies, unguarded sinks, etc.) over an agent skill can then be reduced to Datalog reachability queries. We evaluate Semia on 13,728 real-world skills from public marketplaces. Semia renders all of them auditable and finds that more than half carry at least one critical semantic risk. On a stratified sample of 541 expert-labeled skills, Semia achieves 97.7% recall and an F1 of 90.6%, substantially outperforming signature-based scanners and LLM baselines.

2026-05-01T00:48:47Z Hongbo Wen Ying Li Hanzhi Liu Chaofan Shou Yanju Chen Yuan Tian Yu Feng http://arxiv.org/abs/2601.14252v6 2026-04-30T21:34:19Z

Symbolic systems operate over precise identities: variables denote specific objects, pointers target precise memory locations, and database keys refer to singular records. Neural embeddings generalize by compressing away semantic detail, but this compression creates collision ambiguity: multiple distinct entities can share the same representation value. Exact identity recovery requires additional information precisely when representation fibers have size greater than one. The residual cost is controlled by a single combinatorial object: the collision-fiber geometry of the representation map $π$. Let $A_π=\max_u |π^{-1}(u)|$ be the largest collision fiber. The finite laws include a tight fixed-length converse $L \ge \log_2 A_π$, an exact finite-block scaling law, a pointwise adaptive budget $\lceil \log_2 |π^{-1}(u)|\rceil$, and an exact fiberwise rate-distortion law for arbitrary finite sources via recoverable-mass decomposition across representation fibers. The uniform single-block formula $D^\star(L)=\max(0,1-2^L/a)$ appears as a closed-form special case when all mass lies on one collision block, where $a = A_π$ is the collision block size. The same fiber geometry determines query complexity and canonical structure for distinguishing families. Because this residual ambiguity is structural rather than representation-specific, symbolic identity mechanisms (handles, keys, pointers, nominal tags) are the necessary system-level complement to any non-injective semantic representation. All main results are machine-checked in Lean 4.

2026-01-20T18:58:51Z Main PDF: 12 pages, 1 table. Supplementary: 4 pages, 2 tables. Lean 4 artifact available at https://doi.org/10.5281/zenodo.18123531 Tristan Simas http://arxiv.org/abs/2605.13864v1 2026-04-30T17:27:10Z

GPUs have become essential in modern high performance computing, but programming them correctly remains a significant challenge. This difficulty arises from subtle concurrency bugs that result from the explicit management of synchronization primitives and data movement across intricate hierarchies of memory and parallel threads. At the same time, the ability to control these aspects explicitly is at the core of the performance gains granted by GPUs. These challenges have motivated interest in safe GPU programming: tools and languages that can prevent or detect such bugs statically. However, existing approaches make tradeoffs in three dimensions: the strength of their guarantees, the degree of low-level control they allow, and the amount of additional effort required to achieve these safety guarantees. This thesis presents OptiGPU, a system for GPU programming with strong safety guarantees-data race freedom, deadlock freedom, and full functional correctness-that minimizes tradeoffs in the other two dimensions compared to previous approaches. OptiGPU applies proof-preserving compilation to GPU programming, allowing verification of low-level, optimized GPU programs through refinement of simple, verified CPU programs. An OptiGPU user thus avoids the substantial proof burden of directly verifying complex optimized GPU programs, instead directing this refinement with source-to-source transformations that automatically preserve proofs. OptiGPU is implemented as a set of extensions to OptiTrust, an existing framework for proof-preserving compilation on CPUs. OptiGPU models essential GPU programming features, including kernel launches, shared memory, and synchronous barriers, and produces both device and host-side code. We evaluate OptiGPU on two case studies, matrix transpose and tree-based parallel reduction, showing it can derive CUDA code matching techniques found in handwritten references.

2026-04-30T17:27:10Z Master's thesis Julien de Castelnau Thomas Koehler Arthur Charguéraud Clément Pit-Claudel http://arxiv.org/abs/2509.16248v3 2026-04-30T17:17:15Z

This paper presents GRAPHMEND, a high-level compiler technique that eliminates FX graph breaks in PyTorch 2 programs. Although PyTorch 2 introduced TorchDynamo and TorchInductor to enable just-in-time graph compilation, unresolved dynamic control flow and unsupported Python constructs often fragment models into multiple FX graphs. These fragments force frequent fallbacks to eager mode, introduce costly CPU-to-GPU synchronizations, and reduce optimization opportunities. GRAPHMEND addresses this limitation by analyzing and transforming source code before execution. Built on the Jaseci compilation framework, GRAPHMEND introduces two code transformations that remove graph breaks due to dynamic control flow and Python side effects. This design allows PyTorch's compilation pipeline to capture larger, uninterrupted FX graphs without requiring manual refactoring by developers. Evaluation across eight Hugging Face models shows that GRAPHMEND removes graph breaks due to dynamic control flow and Python side effects, reducing the break count to 0 in 6 models and reducing it from 5 to 2 in another model. On NVIDIA RTX 3090 and A40 GPUs, GRAPHMEND achieves up to 75% latency reductions and up to 8% higher end-to-end throughput. These results demonstrate that high-level code transformation is an effective complement to PyTorch's dynamic JIT compilation pipeline, substantially improving both usability and performance.

2025-09-17T17:15:35Z Savini Kashmira Jayanaka Dantanarayana Thamirawaran Sathiyalogeswaran Krisztian Flautner Lingjia Tang Jason Mars http://arxiv.org/abs/2603.29716v2 2026-04-30T14:25:43Z

We present a graded modal type theory, a dependent type theory with grades that can be used to enforce various properties of the code. The theory has $Π$-types, weak and strong $Σ$-types, natural numbers, an empty type, and a universe, and we also extend the theory with weak and strong unit types and graded $Σ$-types. The theory is parameterized by a modality structure, a kind of partially ordered semiring, whose elements (grades) are used to track the usage of variables in terms and types. Different modalities are possible. We focus mainly on quantitative properties, in particular erasure: with the erasure modality one can mark function arguments as erasable. The theory is fully formalized in Agda. The formalization, which uses a syntactic Kripke logical relation at its core and is based on earlier work, establishes major meta-theoretic properties such as subject reduction, consistency, normalization, and decidability of definitional equality. We also prove a substitution theorem for grade assignment, and preservation of grades under reduction. Furthermore we study an extraction function that translates terms to an untyped $λ$-calculus and removes erasable content, in particular function arguments with the "erasable" grade. For a certain class of modalities we prove that extraction is sound, in the sense that programs of natural number type have the same value before and after extraction. Soundness of extraction holds also for open programs, as long as all variables in the context are erasable, the context is consistent, and erased matches are not allowed for weak $Σ$-types.

2026-03-31T13:16:11Z Andreas Abel Nils Anders Danielsson Oskar Eriksson http://arxiv.org/abs/2604.27863v1 2026-04-30T13:46:44Z

Functional logic languages are a high-level approach to programming by combining the most important declarative features. They abstract from small-step operational details so that programmers can concentrate on the logical aspects of an application. This is supported by appropriate evaluation strategies. Demand-driven evaluation from functional programming is amalgamated with non-determinism from logic programming so that solutions or values are computed whenever they exist. This frees the programmer from considering the influence of an operational strategy on the success of a computation, but it is a challenge to the language implementer. A non-deterministic demand-driven strategy might duplicate unevaluated choices of an expression, which could duplicate the computational effort. In recent implementations, this problem has been tackled by adding a kind of memoization of non-deterministic choices to the expression under evaluation. Since this has been implemented in imperative target languages, it was unclear whether this could also be supported in a functional programming environment like Haskell. This paper presents a solution to this challenge by transforming functional logic programs into a monadic representation. Although this transformation is not new, we present an implementation of the monadic interface which supports memoization in non-deterministic branches. Additionally, we include more advanced features of functional logic languages, namely functional patterns and encapsulated search, in our approach. By optimizing our implementation for purely functional computations with both a static and dynamic approach, we are able to achieve a promising performance that outperforms current compilers for Curry.

2026-04-30T13:46:44Z Under consideration in Theory and Practice of Logic Programming (TPLP) Michael Hanus Kai-Oliver Prott Finn Teegen http://arxiv.org/abs/2602.05528v2 2026-04-30T08:10:28Z

Asynchronous effects of Ahman and Pretnar complement the conventional synchronous treatment of algebraic effects with asynchrony based on decoupling the execution of algebraic operation calls into signalling that an operation's implementation needs to be executed, and into interrupting a running computation with the operation's result, to which the computation can react by installing matching interrupt handlers. Beyond providing asynchrony for algebraic effects, the resulting core calculus also naturally models examples such as pre-emptive multi-threading, (cancellable) remote function calls, and multi-party applications. In this paper, we study the normalisation properties of this calculus. We prove that if one removes general recursion from it, then the remaining calculus is strongly normalising, including both its sequential and parallel parts. To cover more interesting programs, we also prove that the sequential part of the calculus remains strongly normalising when a controlled amount of interrupt-driven recursive behaviour is reintroduced. Our normalisation proofs are structured compositionally as an extension of Lindley and Stark's $\top\top$-lifting-based approach for proving strong normalisation of effectful languages. All our results are also formalised in Agda.

2026-02-05T10:39:09Z FSCD 2026 final version Danel Ahman Ilja Sobolev http://arxiv.org/abs/2604.14942v3 2026-04-30T06:14:41Z

Programming is about automation in a wide variety of domains. Developing itself is one of those. As a side-effect, progress in automated coding may make people less willing to learn computer programming. This could become an issue, if the skill of computational problem solving is not only for the immediate economic benefit, but an important part of our knowledge about the world. We suggest that weakened incentives can be countered by lowering the entry barrier. We plan to shorten learning time by reducing the accidental complexity of the programming language and its runtime system. We describe a session plan that introduces programming and computing fundamentals for novices, assuming only basic mathematical background. This requires a non-mainstream, functional and concatenative language. This language, CON-CAT, is a by-product of research in category theory. It provides direct access to fundamental ideas like recursion and advanced ones like Gödel-encoding in an entertaining puzzle-like manner.

2026-04-16T12:36:44Z 6 pages, 1 figure, v3 small corrections and additional references, final version will be published elsewhere Attila Egri-Nagy http://arxiv.org/abs/2604.27268v1 2026-04-29T23:48:07Z

Behavioural distances provide a quantitative approach to comparing the states of transition systems, moving beyond traditional Boolean notions of equivalence. In this paper, we develop a sound and complete axiomatisation of behavioural distance for nondeterministic processes using Milner's charts, a model that generalises finite-state automata by incorporating variable outputs. Charts provide a compelling setting for studying behavioural distances because they shift the focus from language equivalence to bisimilarity. Their axiomatic study lays the groundwork for quantitative analysis of more expressive models, such as weighted transition systems. To formalise this approach, we adopt string diagrams as our syntax of choice. String diagrams closely mirror the graphical structure of charts, while providing a rigorous formalism that supports inductive reasoning and compositional semantics. Unlike traditional algebraic syntaxes, which require additional mechanisms such as binders and substitution, string diagrams offer a variable-free representation where recursion naturally decomposes into simpler components. This makes them well-suited for reasoning about behavioural distances and aligns with broader efforts to axiomatise automata-theoretic equivalences through a unified diagrammatic framework.

2026-04-29T23:48:07Z Accepted to ICALP 2026 Wojciech Różowski Robin Piedeleu Alexandra Silva Fabio Zanasi http://arxiv.org/abs/2604.07626v2 2026-04-29T22:12:58Z

Token identity is semantic information for measurement-bearing expressions. Intervals, dimension tags, and token-erased syntax can say what values a measured leaf may take, but they cannot say whether two occurrences name the same observation or two fresh observations. We give a small formal semantics in which each measured leaf carries an interval of possible exact values and an opaque observation-event token. Here "token" means an identity for a measurement event, not a lexical token of the source syntax. The denotation of an expression is its warranted enclosure: the set of exact values still justified by hidden-value environments that assign one value to each observation token and respect the declared intervals. Over this semantics, e -> e' is a claim-tightening judgment, equivalently enclosure containment Encl(e') subseteq Encl(e), while interchangeability is equality of enclosures. The distinction is visible in cancellation, background subtraction, and self-division: reusing one token gives interchangeability with the expected simplified expression, while using distinct tokens gives only one-way containment. We prove that provenance-blind summaries of the kind studied here, preserving intervals, dimension tags, and token-erased syntax, are insufficient to recover the correct rewrite class. The formal results are mechanized in Lean 4 with no sorry or admit placeholders.

2026-04-08T21:58:01Z 17 pages; Lean 4 formalization; prepared for submission to South American Journal of Logic David B. Hulak Arthur F. Ramos Ruy J. G. B. de Queiroz http://arxiv.org/abs/2604.09961v2 2026-04-29T14:45:48Z

Dominance is a fundamental concept in compilers based on static single assignment (SSA) form. It underpins a wide range of analyses and transformations and defines a core property of SSA: every use must be dominated by its definition. We argue that this reliance on dominance has become increasingly problematic -- both in terms of precision and applicability to modern higher-order languages. First, control flow overapproximates data flow, which makes dominance-based analyses inherently imprecise. Second, dominance is well-defined only for first-order control-flow graphs (CFGs). More critically, higher-order programs violate the assumptions underlying SSA and classic CFGs: without an explicit CFG, the very notion that all uses of a variable must be dominated by its definition loses meaning. We propose an alternative foundation based on free variables. In this view, $φ$-functions and function parameters directly express data dependencies, enabling analyses traditionally built on dominance while improving precision and naturally extending to higher-order programs. We further present an efficient technique for maintaining free-variable sets in a mutable intermediate representation (IR). For analyses requiring additional structure, we introduce the nesting tree -- a relaxed analogue of the dominator tree constructed from variable dependencies rather than control flow. Our benchmarks demonstrate that the algorithms and data structures presented in this paper scale log-linearly with program size in practice.

2026-04-10T23:54:00Z PLDI 2026 Roland Leißa Johannes Griebler 10.1145/3808286 http://arxiv.org/abs/2604.17612v2 2026-04-29T12:44:22Z

Multi-agent systems built on large language models (LLMs) are difficult to reason about. Coordination errors such as deadlocks or type-mismatched messages are often hard to detect through testing. We introduce a domain-specific language for specifying agent coordination based on message sequence charts (MSCs). The language separates message-passing structure from LLM actions, whose outputs remain unpredictable. We define the syntax and semantics of the language and present a syntax-directed projection that generates deadlock-free local agent programs from global coordination specifications. We illustrate the approach with a diagnosis consensus protocol and show how coordination properties can be established independently of LLM nondeterminism. We also describe a runtime planning extension in which an LLM dynamically generates a coordination workflow for which the same structural guarantees apply. An open-source Python implementation of our framework is available as ZipperGen.

2026-04-19T20:54:30Z 40 pages; v2: All definitions and results are now mechanically verified in Lean 4 Benedikt Bollig Matthias Függer Thomas Nowak http://arxiv.org/abs/2601.01638v2 2026-04-29T09:11:49Z

The relational semantics of linear logic is a powerful framework for defining resource-aware models of the $λ$-calculus. However, its quantitative aspects are not reflected in the preorders and equational theories induced by these models. Indeed, they can be characterized in terms of (in)equalities between Böhm trees up to extensionality, which are qualitative in nature. We employ the recently introduced checkers calculus to provide a quantitative and contextual interpretation of the preorder associated to the relational semantics. This way, we show that the relational semantics refines the contextual preorder constraining the number of interactions between the related terms and the context.

2026-01-04T19:04:05Z Version with proofs of the FoSSaCS 2026 paper Adrienne Lancelot Giulio Manzonetto Guy McCusker Gabriele Vanoni