arXiv subjects

Adithya Murali

Publications and source records attributed to Adithya Murali.

15 recordsLinked to original sources

arXiv2026

Faithful Autoformalization of Natural Language Assertions

Formal contracts are essential for software testing and verification, yet writing them remains labor-intensive and error-prone. LLMs offer a promising path toward autoformalization: synthesizing executable assertions from natural-language specifications and thereby bridging the gap between informal developer intent and formal executable specifications. We present Monty: an autoformalization framework for assertions that tackles the challenges of expectations of validity of assertions and ambiguity in natural-language. Our techniques are based on filtering formalizations using a novel conformance score metric and validity scores obtained from testing the code against formalized assertions. We evaluate our approach on 541 assertion-generation tasks derived from 22 collection-like Java classes, and show that our technique produces the ground truth more reliably (improving upto 20 points in precision on average) than when using LLMs naively to translate assertions.

cs.SE↗

arXiv2026

Imprompt: A Language Framework for Prompt Programming

With the unprecedented success of Language Models (LMs), the science of Prompt Engineering has evolved the powerful idea of Prompt Programming, where prompts are treated as a programmable control surface for describing complex tasks and leveraging LM capabilities. However, existing prompt programming frameworks suffer from various complexities and inelegances, which make them hard to utilize in practice for effectively describing tasks. We propose Imprompt, a new language framework for the study and practice of prompt programming. We undertake a foundational investigation of prompt programming, and contend that prompt programs must contain only the task descriptions and must be decoupled from lower-level 'execution' details. We further develop this position by illustrating structured prompting as a combination of prompt programming and prompt program 'compilation'. We exemplify this view by formally defining two compilers for Imprompt programs. We then explore the idea of typing for prompt programs and draw a correspondence between type checking and constrained decoding. Finally, we implement our compilers and type checkers and evaluate them on a variety of case studies. We believe our work contributes programming-language foundations toward the emerging area of prompt programming.

cs.AI↗

arXiv2026

Programmers Are Poor and Overconfident Judges of LLM-Generated Assertions

Code comprehension and code review are already critically important software engineering tasks, and the rising use of AI code generation tools is only increasing that importance. Generative AI has the possibility of supporting these activities, for example by augmenting code with assertions and natural-language explanations describing code behavior. However, little is known about how effective such support may be. We conduct a controlled experiment with 86 Python programmers and a follow-up think-aloud study to examine developers' ability to assess the correctness and completeness of generated assertions of varying quality, and to investigate how natural-language explanations influence these assessments. While programmers can somewhat accurately judge correct assertions (74% accuracy), they perform poorly when shown incorrect assertions (49% accuracy), despite reporting similar levels of confidence in both judgments. This difference in judgment accuracy is statistically significant (p < 0.001): the odds of a developer accurately judging a correct assertion was nearly three times higher than the odds of accurately judging an incorrect assertion (OR = 2.94). Surprisingly, natural-language explanations of assertions provided no overall benefit. Furthermore, low-quality explanations could impair specification assessment accuracy (p = 0.037, OR = 0.58) while simultaneously increasing developer confidence (p = 0.005, 3.99/5 vs. 4.25/5). Our findings suggest that, contrary to common assumptions, AI assistance may not improve the reliability of code comprehension and review. More broadly, our findings highlight the importance of helping developers evaluate machine-generated reliability artifacts, in addition to generating them.

cs.SE↗

arXiv2026

Counterexample Guided Learning in the Large using Reasoning Agents

LLMs and LLM agents should improve when given feedback, but identifying when they are able to do so is difficult: feedback is heterogeneous, domain-specific, and difficult to control. We approach this challenge by asking LLMs to perform regular-expression induction, a classical symbolic learning problem where precise mechanisms for feedback exist in the form of counterexamples. In counterexample-guided learning, a learner (LLM) proposes candidate regular expressions from positive/negative-labeled strings, and the teacher (verifier) returns counterexamples showcasing the difference between the candidate and target languages. We identify novel counterexample-guided refinement strategies that enable effective regex learning, such as regularization and symbolic counterexample clusters. We also explore agentic strategies such as reflection and repair loops. Empirically, we find that verifier feedback substantially improves sample efficiency on challenging regex-induction tasks, reducing the number of labeled examples required and enabling learning of complex target expressions where standard prompting fails. For example, on the hardest task groups, our counterexample-guided framework improves success from 3.2% to 38.1% and from 38.9% to 74.1% on two different regex domains. These results suggest that LLMs can benefit from rich feedback beyond treating it as additional data, opening the door for robust verifier-guided methods for LLM-based program synthesis and formal reasoning.

cs.LG↗

arXiv2026

Complete first-order reasoning for functional programs

Several practical tools for automatically verifying functional programs (e.g., Liquid Haskell and Leon for Scala programs) rely on a heuristic based on unrolling recursive function definitions followed by quantifier-free reasoning using SMT solvers. We uncover foundational theoretical properties of this heuristic, revealing that it can be generalized and formalized as a technique that is in fact complete for reasoning with combined First-Order theories of algebraic datatypes and background theories, where background theories support decidable quantifier-free reasoning. The theory developed in this paper explains the efficacy of these heuristics when they succeed, explains why they fail when they fail, and the precise role that user help plays in making proofs succeed.

cs.LO↗

arXiv2026

Verification Modulo Tested Library Contracts

We consider the problem of verification modulo tested library contracts as a step towards automating the verification of client programs that use complex libraries. We formulate this problem as the synthesis of modular contracts for the library methods used by the client that are adequate to prove the client correct, and that also pass the scrutiny of a testing engine that tests the library against these contracts. We also consider a new form of method contracts called contextual contracts that arise in this setting that hold in the context of the client program, and can often be simpler and easier to infer than classical modular contracts. We provide a counterexample-guided learning framework to solve this problem, in which the synthesizer interacts with a constraint solver as well as the testing engine in order to infer adequate modular/contextual method contracts and inductive invariants for the client. The main synthesis engines we use are generalizing CHC solvers that are realized using ICE learning algorithms. We realize this framework in a tool called DUALIS and show its efficacy on benchmarks where clients call large libraries.

cs.PL↗

arXiv2026

FO-Complete Program Verification for Heap Logics

We develop the first two heap logics that have implicit heaplets and that admit FO-complete program verification. The notion of FO-completeness is a theoretical guarantee that all theorems that are valid when recursive definitions are interpreted as fixpoint definitions (instead of least fixpoint) are guaranteed to be eventually proven by the system. The logics we develop are a frame logic ($\textit{FL}$) and a separation logic ($\textit{SL-FL}$) that has an alternate semantics inspired by frame logic. We show verification condition generation for FL that is amenable to FO-complete reasoning using quantifier instantiation and SMT solvers. We show $\textit{SL-FL}$ can be translated to FL in order to obtain FO-complete reasoning. We implement tools that realize our technique and show the expressiveness of our logics and the efficacy of the verification technique on a suite of benchmarks that manipulate data structures.

cs.LO↗

arXiv2025

Separating the Wheat from the Chaff: Understanding (In-)Completeness of Proof Mechanisms for Separation Logic with Inductive Definitions

For over two decades Separation Logic has been arguably the most popular framework for reasoning about heap-manipulating programs, as well as reasoning about shared resources and permissions. Separation Logic is often extended to include inductively-defined predicates, interpreted as least fixpoints, forming Separation Logic with Inductive Definitions (SLID). Many theoretical and practical advances have been made in developing automated proof mechanisms for SLID, but these mechanisms are imperfect, and a deeper understanding of their failures is desired. As expressive as Separation Logic is, it is not surprising that it is incomplete, and in fact, it contains several sources of incompleteness that defy automated reasoning. In this paper we study these sources of incompleteness and how they relate to failures of proof mechanisms. We place SLID within a larger logic, that we call Weak Separation Logic (WSL). We prove that unlike SLID, WSL is complete for a non-trivial fragment of quantified entailments with background theories and inductive definitions, via a reduction to first-order logic (FOL). Moreover, we show that the ubiquitous fold/unfold proof mechanism is sound and complete for theory-free, quantifier-free WSL entailments with inductive definitions. Through this, we understand proof failures as stemming from nonstandard models present in WSL, but not allowed in SLID. These rogue models are typically infinite, and we use the formalism of symbolic structures to represent and automatically find them. We present a prototype tool that implements the FOL encoding of WSL and test it on an existing benchmark, which contains over 700 quantified entailment problems with inductive definitions. Our tool is able to find counter-models to many of the examples, and we provide a partial taxonomy of the rogue models, shedding some light on real-world proof failures.

cs.LO↗

arXiv2024

Predictable Verification using Intrinsic Definitions

We propose a novel mechanism of defining data structures using intrinsic definitions that avoids recursion and instead utilizes monadic maps satisfying local conditions. We show that intrinsic definitions are a powerful mechanism that can capture a variety of data structures naturally. We show that they also enable a predictable verification methodology that allows engineers to write ghost code to update monadic maps and perform verification using reduction to decidable logics. We evaluate our methodology using Boogie and prove a suite of data structure manipulating programs correct.

cs.PL↗

arXiv2022

Applying Deep Reinforcement Learning to the HP Model for Protein Structure Prediction

A central problem in computational biophysics is protein structure prediction, i.e., finding the optimal folding of a given amino acid sequence. This problem has been studied in a classical abstract model, the HP model, where the protein is modeled as a sequence of H (hydrophobic) and P (polar) amino acids on a lattice. The objective is to find conformations maximizing H-H contacts. It is known that even in this reduced setting, the problem is intractable (NP-hard). In this work, we apply deep reinforcement learning (DRL) to the two-dimensional HP model. We can obtain the conformations of best known energies for benchmark HP sequences with lengths from 20 to 50. Our DRL is based on a deep Q-network (DQN). We find that a DQN based on long short-term memory (LSTM) architecture greatly enhances the RL learning ability and significantly improves the search process. DRL can sample the state space efficiently, without the need of manual heuristics. Experimentally we show that it can find multiple distinct best-known solutions per trial. This study demonstrates the effectiveness of deep reinforcement learning in the HP model for protein folding.

cs.LG↗

arXiv2020

Model-Guided Synthesis of Inductive Lemmas for FOL with Least Fixpoints

Recursively defined linked data structures embedded in a pointer-based heap and their properties are naturally expressed in pure first-order logic with least fixpoint definitions (FO+lfp) with background theories. Such logics, unlike pure first-order logic, do not admit even complete procedures. In this paper, we undertake a novel approach for synthesizing inductive hypotheses to prove validity in this logic. The idea is to utilize several kinds of finite first-order models as counterexamples that capture the non-provability and invalidity of formulas to guide the search for inductive hypotheses. We implement our procedures and evaluate them extensively over theorems involving heap data structures that require inductive proofs and demonstrate the effectiveness of our methodology.

cs.LO↗

arXiv2020

Decidable Synthesis of Programs with Uninterpreted Functions

We identify a decidable synthesis problem for a class of programs of unbounded size with conditionals and iteration that work over infinite data domains. The programs in our class use uninterpreted functions and relations, and abide by a restriction called coherence that was recently identified to yield decidable verification. We formulate a powerful grammar-restricted (syntax-guided) synthesis problem for coherent uninterpreted programs, and we show the problem to be decidable, identify its precise complexity, and also study several variants of the problem.

cs.PL↗

arXiv2019

Deciding Memory Safety for Single-Pass Heap-Manipulating Programs

We investigate the decidability of automatic program verification for programs that manipulate heaps, and in particular, decision procedures for proving memory safety for them. We extend recent work that identified a decidable subclass of uninterpreted programs to a class of alias-aware programs that can update maps. We apply this theory to develop verification algorithms for memory safety--- determining if a heap-manipulating program that allocates and frees memory locations and manipulates heap pointers does not dereference an unallocated memory location. We show that this problem is decidable when the initial allocated heap forms a forest data-structure and when programs are streaming-coherent, which intuitively restricts programs to make a single pass over a data-structure. Our experimental evaluation on a set of library routines that manipulate forest data-structures shows that common single-pass algorithms on data-structures often fall in the decidable class, and that our decision procedure is efficient in verifying them.

cs.PL↗

arXiv2019

Composing Neural Learning and Symbolic Reasoning with an Application to Visual Discrimination

We consider the problem of combining machine learning models to perform higher-level cognitive tasks with clear specifications. We propose the novel problem of Visual Discrimination Puzzles (VDP) that requires finding interpretable discriminators that classify images according to a logical specification. Humans can solve these puzzles with ease and they give robust, verifiable, and interpretable discriminators as answers. We propose a compositional neurosymbolic framework that combines a neural network to detect objects and relationships with a symbolic learner that finds interpretable discriminators. We create large classes of VDP datasets involving natural and artificial images and show that our neurosymbolic framework performs favorably compared to several purely neural approaches.

cs.LG↗

arXiv2019

A First-Order Logic with Frames

We propose a novel logic, called Frame Logic (FL), that extends first-order logic (with recursive definitions) using a construct Sp(.) that captures the implicit supports of formulas -- the precise subset of the universe upon which their meaning depends. Using such supports, we formulate proof rules that facilitate frame reasoning elegantly when the underlying model undergoes change. We show that the logic is expressive by capturing several data-structures and also exhibit a translation from a precise fragment of separation logic to frame logic. Finally, we design a program logic based on frame logic for reasoning with programs that dynamically update heaps that facilitates local specifications and frame reasoning. This program logic consists of both localized proof rules as well as rules that derive the weakest tightest preconditions in FL.

cs.LO↗