arXiv cs.PL Papers¶
Recent papers from arXiv Programming Languages category.
2026¶
A Domain-Specific Language for Programming Non-Axiomatic Reasoning Agents¶
Authors: Seamus Brady et al.
arXiv: 2604.00043 | PDF
Date: March 29, 2026
Topics: DSL, non-axiomatic reasoning, NARS, agent programming
Non-Axiomatic Reasoning Systems (NARS) provide a framework for building adaptive agents that operate under insufficient knowledge and resources. However, the standard input language, Narsese, poses a usability barrier with its dense symbolic notation and overloaded punctuation.
DriftScript is a Lisp-like domain-specific language that compiles to Narsese. It provides: - Source-level constructs for NAL levels 1-8 - Readable keyword-based S-expressions instead of symbolic syntax - Coverage of inheritance, temporal implication, variable quantification, sequential conjunction, and operation invocation
Implementation: - Zero-dependency, four-stage compiler pipeline - 1,941 lines of C99 - Connects to external systems through structured callbacks and HTTP operation registry - Enables sense-reason-act loop for autonomous agents
Evaluation: 106-case test suite, equivalence testing against hand-written Narsese, NAL coverage analysis, structural readability metrics, and compilation benchmarks.
Executing as You Generate: Hiding Execution Latency in LLM Code Generation¶
Authors: Zhensu Sun et al.
arXiv: 2604.00491 | PDF
Date: April 1, 2026
Topics: LLM code generation, parallel execution, latency optimization
Current LLM-based coding agents follow a serial execution paradigm: generate complete code, then execute it. This leaves the executor idle during generation and the generator idle during execution. This paper observes that LLMs produce code tokens sequentially without revision, making parallel execution possible.
Eager is a concrete implementation featuring: - AST-based chunking - Dynamic batching with gated execution - Early error interruption
Results: Reduces non-overlapped execution latency by up to 99.9% and end-to-end latency by up to 55% across seven LLMs and four benchmarks.
Key contributions: - Formalizes parallel execution paradigm as a three-stage pipeline (generation, detection, execution) - Derives closed-form latency bounds - Demonstrates significant speedup in practice
Superset Decompilation¶
Authors: Chang Liu et al.
arXiv: 2603.28002 | PDF
Date: March 30, 2026
Topics: reverse engineering, decompilation, Datalog, program analysis
Reverse engineering tools remain monolithic and imperative compared to the advancement of modern compiler architectures. This paper formalizes provenance-guided superset decompilation (PGSD), a framework that monotonically derives facts about binaries into a relation store. Instead of committing early to a single interpretation, the pipeline retains ambiguous interpretations as parallel candidates with provenance, deferring resolution until the final selection phase.
Manifold implements PGSD as a declarative reverse engineering framework that lifts Linux ELF binaries to C99 through a granular intermediate representation in ~35K lines of Rust and Datalog. On GNU coreutils, Manifold's output quality matches Ghidra, IDA Pro, angr, and RetDec while producing fewer compiler errors.
Key contributions: - Treats decompilation as the reverse of compilation with modular passes - Uses Datalog for declarative binary analysis - Maintains multiple interpretations with provenance tracking - Generalizes across compilers and optimization levels
Towards verifying unsafe Rust programs against Rust's pointer-aliasing restrictions¶
Authors: Wannes Tas et al.
arXiv: 2603.28326 | PDF
Date: March 30, 2026
Topics: Rust, formal verification, program logics, pointer aliasing
The Rust programming language enforces strong ownership rules (exclusive ownership, mutable references, or shared references) through its type system, except inside unsafe blocks. These pointer-aliasing rules are exploited by the compiler for optimization but break down in unsafe code.
This paper presents work in progress towards the first program logic for modularly verifying that Rust programs using unsafe blocks comply with the pointer-aliasing rules. This addresses a critical gap in Rust's safety story - ensuring that unsafe code doesn't violate the assumptions that safe code relies on.
Key contributions: - First program logic for verifying unsafe Rust against aliasing rules - Modular verification approach - Bridges the gap between Rust's type system and unsafe code
Multi-paradigm Logic Programming in the ErgoAI System¶
Authors: Theresa Swift et al.
arXiv: 2603.29819 | PDF
Date: March 31, 2026
Topics: logic programming, F-logic, HiLog, well-founded semantics
ErgoAI is a high-level, multi-paradigm logic programming language and system developed as a successor to Flora-2. It's oriented towards scalable knowledge representation and reasoning, exploiting both structured knowledge and knowledge from external sources like vector embeddings.
Key features: - Well-founded semantics for reasoning - Object-based logic (F-logic) with non-monotonic inheritance - Higher-order syntax (HiLog style) - Defeasibility of rules - Semantically clean transactional updates - Subgoal delay for handling unsafe queries - Optional bounded rationality at module level
Although Flora-2 programs compile to XSB and adopt Prolog features, ErgoAI is altogether a different language and system.
Folding the Heighway dragon curve¶
Authors: Shin-Cheng Mu
arXiv: 2603.27302 | PDF
Date: March 28, 2026
Topics: functional programming, fold duality, fractals
The Heighway dragon curve is a well-known fractal with two construction methods: (1) repeatedly copy, rotate 90°, and connect; (2) repeatedly replace each segment with two segments at a right angle. This paper proves their equivalence using functional programming theory.
Key insight: The two construction approaches are respectively a foldr and a foldl, and their equivalence follows from the second duality theorem with the distributivity of an "interleave" operator.
The paper generalizes the construction to allow rotations to both sides, making the connection to fold duality clearer.
Key contributions: - Formal proof of equivalence using fold duality - Application of functional programming theory to fractals - Generalization with bidirectional rotations