Coding agents change running software: they patch a service's code or overwrite its stored state, and then act on their own expectation of how the service will respond afterwards. A wrong expectation may surface only several calls later. Function-level code-execution benchmarks omit persistent service state, and agent benchmarks score the actions an agent takes or the final state it reaches. We introduce CTE-Bench, which measures whether a model can predict how an intervention changes a stateful service's future behavior, without asking it to choose actions. Each scenario gives the model Python service code, the calls and responses observed before the intervention, the intervention itself (a source edit or a state overwrite), and 40 fixed future calls; the model predicts every future response, and predictions are checked by executing the service. Three memory protocols control whether the model sees the correct earlier responses, none of them, or its own earlier predictions. CTE-Bench-Core-v1 contains 255 scenarios over six deterministic Python services, giving 10,200 predictions per model. The main score is effect-step value match (VM): exact response equality on the 2,476 future calls whose response the intervention changes. With correct earlier responses revealed, four API-hosted models (DeepSeek V4-Flash, Kimi K2.5, Qwen3.6-35B-A3B, and Claude Sonnet 4.6) reach 54.3%-61.7% effect-step VM. Hiding those responses lowers effect-step VM to 23.2%-29.0%; conditioning on self-generated predictions gives 24.8%-33.2%, and at most 1.2% of scenarios are predicted exactly end to end. Current models thus track intervention effects mainly when correct feedback is supplied, and their errors compound over a rollout. We release CTE-Bench-Core-v1 with its executable oracle, evaluation scripts, and an evaluation card mapping each claim to its protocol.
Kshitish Ghate, Michael Saxon, Tadayoshi Kohno, Aylin Caliskan
Large language models (LLMs) are often conditioned into specific roles such as tutors, interviewers, or agents acting for an organisation, through system prompts that specify how they should behave. The extent to which they can actually adhere to these role specifications under multi-turn interactions with realistic and potentially problematic users is understudied. To address this, we propose user behaviour fuzzing, a guided search over plausible user behaviours that can lead a model to violate its role specification, and introduce RoleFUZZER, a framework that adapts coverage-guided fuzzing from software testing to operationalise the search. RoleFUZZER constructs scenarios guided by a user-behaviour coverage grid that crosses a grounded taxonomy of user interaction types and the role specification of the model under test i.e. what is under pressure. With RoleFUZZER, we evaluate LLMs across five role specifications, including tutor, financial advisor, AI interviewer, antagonistic roleplay character, and web agent under organisational policies. Across simulated and validated conversations in 15 campaigns, the framework reaches >= 89% coverage of the 64-cell behaviour grid for every model and role tested, and surfaces specification-grounded failure patterns. At an equal budget of conversations, removing the RoleFUZZER grid and asking an LLM to generate specification targeting users free-form lowers coverage to 39-69% with fewer kinds of violations. The same specification can also fail in opposite directions. For e.g., on the antagonist, open-weight target models override the player's safety signals, while Claude Haiku 4.5 breaks character instead. RoleFUZZER offers a tractable way to uncover vulnerabilities in LLMs across role specifications by testing diverse user behaviours that are likely to violate them.
Ruopei Sun, Jianfeng Cai, Kangwen Zhao, Wengang Zhou, Houqiang Li
Context compaction lets long-horizon coding agents continue beyond their context windows, but can erase the state needed to finish a task. Prior evaluations measure efficiency, retained information, or end-to-end success, leaving the effect of a specific history replacement difficult to isolate. Compact-Bench pairs full-history and summary-only continuations from a shared repository state, and evaluates recurring details and successive production compactions. It covers 35 SWE-bench repairs, 146 agentic sessions, and 106 production chains. Across 14 summarizer backbones, stable damage ranges from 5-64% of 22 full-history successes: Claude Opus 4.8 preserves the most repairs, while DeepSeek-Flash damages the most. With Kimi-K3 as summarizer, the Kimi Code prompt preserves the most task continuations (19/22), whereas Claude Code retains the most recurring details (15.0%). LLM summaries retain at most 16.0% of recurring details, compared with 79.1% for mechanical extraction; production summaries retain 15.7% at the first boundary. Failures recur across methods: one repair is damaged by 13/14 backbones, and another by all 5 prompts. These results show that reliable compaction requires preserving the progress needed to complete a task, and that retaining relevant information alone does not ensure successful continuation. Code is available at https://github.com/cai-jianfeng/compact-bench and data is available at https://huggingface.co/datasets/caijanfeng/compact-bench.
Serverless computing, particularly Function-as-a-Service (FaaS), has revolutionized cloud computing by abstracting infrastructure management and enabling dynamic resource allocation. This paper examines the performance and compatibility of OpenFaaS, an open-source serverless platform, when deployed on various Kubernetes distributions, including Kubeadm, K3s, MicroK8s, and K0s. Moreover, leveraging the CloudLab infrastructure, this study examines the impact of Python, Go, and Node$.$js programming languages on the performance of Kubernetes-enabled OpenFaaS, specifically when these languages are used to develop functions deployed on the platform. The performance is evaluated and analyzed under various levels of concurrent invocations using several usage-level metrics, such as throughput and CPU usage, as well as responsiveness metrics, such as delay. According to our findings, Go consistently outperforms Python and Node$.$js in terms of throughput and CPU usage, making it the ideal runtime for serverless applications. Among the Kubernetes distributions, K3s and Kubeadm exhibit superior performance, with Kubeadm maintaining low latency and efficient CPU usage, and K3s demonstrating high throughput. This study provides valuable insights into optimizing the Kubernetes-enabled OpenFaaS platform, highlighting the strengths and trade-offs of different Kubernetes distributions and language runtimes.
A range of software engineering tasks, such as optimization and refactoring, seek to preserve the behavior of software components and require methods to assess functional similarity. Research has explored formal approaches to prove functional equivalence and used machine learning methods to approximate it, but the former does not scale and does not quantify the extent of similarity and the latter struggles when subtle semantic differences are present. Differential fuzzing offers a scalable approach for differencing programs semantically, but applying it often requires substantial manual effort to construct input generators that are well-suited to a target pair of programs. To address this challenge, we introduce DIFUZZI, which incorporates an LLM-based generator synthesis step that is iteratively validated and repaired. By automating this step, DIFUZZI paves the way for broader applicability of differential fuzzing. We release DIFUZZI at https://github.com/Dristi123/difuzzi_tool, with a demonstration video at https://youtu.be/FlHNOHuo9WE
Liyan Huang, Kaicheng Wang, Amy Zhang, Weihang Wang
WebAssembly (Wasm) promises seamless reuse of C/C++ codebases as portable, efficient, sandboxed binaries for execution on the Web. In practice, however, this promise often falls short: recent studies show that cross-compiling C/C++ source to Wasm and native binaries frequently produces runtime discrepancies due to differences in library implementation and compiler bugs. Since the root causes are platform-level and hidden beneath the source code, even state-of-the-art LLM-based coding agents can struggle to diagnose and repair these discrepancies. In this paper, we present WasmMend, the first system for automatically repairing Native-Wasm functional discrepancies. WasmMend transforms open-ended debugging into a two-stage process. First, a novel differential trace analysis approach localizes the function where Wasm and native executions first diverge. Second, guided by this localization, a multi-agent system coordinates separate agents to interpret platform-specific behavior and synthesizes patch plans. Experiments on discrepancies from real-world C/C++ projects show that, across five backbone models, WasmMend achieves a fix rate of up to 84.8%, with average gains of at least 37.2 percentage points over the evaluated open-source coding agent baselines, while maintaining comparable cost.
LLM-generated explanations help developers understand automated program repair (APR) rationale and assess patch quality. Incorrect explanations can mislead developers about a repair's reliability, making their detection important. However, manually inspecting the growing volume of generated repairs is impractical, motivating automated detection. Existing methods mainly target incorrect answers from general question-answering LLMs, often by sampling multiple responses and detecting contradictions. Applied to APR, they face two limitations: (1) they may depend on external verifiers that are unavailable or lack APR-specific reasoning; and (2) sampling strategies such as paraphrasing or re-prompting the same input often yield superficial, redundant responses that repeat the original error rather than expose it. We propose MetaReJ, the first method designed to detect incorrect APR explanations using metamorphic testing. Inspired by debate and viva examinations, MetaReJ challenges explanations through an argumentative strategy without requiring external verifiers. It applies novel metamorphic relations (MRs) to recursively question the LLM under equivalent and non-equivalent repair scenarios and detect contradictions. To obtain diverse samples, MetaReJ transforms the buggy program into variants that preserve or alter the underlying bug, then asks the LLM how to repair each variant. Contradictions among responses are encoded as MR violations, which indicate that the original explanation is incorrect. We evaluate MetaReJ on a manually labelled dataset of 1,756 explanations from multiple LLM-based APR systems. MetaReJ achieves an MCC of up to 0.43, outperforming prior methods by up to 333.3% in MCC.
The rise of Large Language Models (LLMs) has significantly advanced text-to-SQL capabilities, but generated queries may have errors. Test datasets are important for catching errors, but having only one or a few test datasets leaves many bugs undetected. Effective testing requires multiple test datasets that can expose subtle semantic bugs and edge cases. Recent work on test data generation and equivalence checking has focused on using constraint solvers. However, constraint-based approaches are slow and not scalable. In this paper, we introduce LASSO, a framework inspired by XData, but designed to minimize the use of constraint solvers, instead prioritizing use of data generation strategies based on fast heuristics. LASSO deploys a number of "attack strategies", each targeting a specific class of semantic errors. LASSO code was built using LLM-based agentic coding. We used carefully designed workflows and rules to support effective code generation and to avoid common code generation issues. A large number of internal test cases are included in LASSO to catch errors in generated code, and more are added when we extend LASSO, to avoid subsequent regressions. Our empirical evaluation across multiple benchmarks demonstrates that LASSO outperforms the state of the art, including XData, SpotIT, ParSEval, and Polygon.
Existing code reveals past implementation decisions but rarely explains the reasoning behind them. When coding agents implement new features or modify existing functionality, past decisions may conflict with new task instructions. This raises a fundamental question for agentic software evolution: when multiple implementations satisfy the same functional requirements, how do coding agents decide whether to follow existing code or new task instructions? We investigate this question through bidirectional, controlled experiments in purpose-built repositories, followed by validation in real-world open-source repositories. Our approach isolates three competing sources of guidance: (1) agents' default preferences, (2) repository precedents, and (3) explicit task instructions. We conduct 5,625 trials across four coding agents and six implementation pairs, with validation in four open-source repositories spanning three programming languages. We find that agents share the same default implementation choices, yet overwhelmingly follow repository precedent when it conflicts with those defaults (359/360 trials). Crucially, not all precedents carry equal influence: three supporting local files yield 96% adoption, compared with only 4% for sibling-resource files, and changing file placement alone can shift adoption from 0% to 100% at a fixed example count. When new task intent conflicts with existing code, written priority rules achieve substantially higher success than code examples alone (59/60 versus 38/60 joint successes for two agents on two implementation pairs). The effectiveness of explicit rules also extends to open-source repositories, where a harness-loaded rule raises joint success from 0/120 to 117/120. These findings motivate design-decision traceability that preserves not only implementation choices but also the rationale and priorities that should guide future changes.
Sparsh Gupta, Kamalavasan Kamalakkannan, Maxim Moraru, Galen Shipman, Patrick Diehl
Scientific applications still rely heavily on legacy Fortran codebases built for homogeneous, CPU-based systems. As High-Performance Computing (HPC) shifts toward GPU-accelerated architectures, many accelerators lack native Fortran bindings, creating a need to translate and optimize legacy code for portability. Kokkos offers performance portability through a single-source C++ abstraction, but manual porting from Fortran demands significant expertise and time. While large language models (LLMs) show promise in source-to-source code generation, their use in autonomous workflows for translating and optimizing parallel code is largely unexplored. This paper presents an agentic AI workflow in which specialized LLM agents collaborate to translate, validate, compile, run, test, debug, and optimize Fortran kernels into portable Kokkos C++ programs without human intervention, bounded by configurable retry and optimization thresholds. We evaluate the pipeline over five runs per configuration on an NVIDIA A100, characterizing its variability in success rates, token cost, and optimization behavior, on benchmark kernels including a real-world multi-file kernel from a production climate model. Proprietary models such as GPT-5 and o4-mini-high completed the full workflow most reliably, producing optimized code that exceeded the Fortran baselines for only a few U.S. dollars, whereas open-weights models such as gpt-oss-120b and Qwen3.5-397B-A17B handled simpler kernels but struggled with complex Kokkos abstractions such as parallel reductions and hierarchical parallelism. This work demonstrates, at the kernel level, the feasibility of agentic AI for Fortran-to-Kokkos transformation and, more broadly, for structured software-engineering tasks such as automated code modernization and refactoring.
Aditya Golatkar, Alessandro Achille, Renos Zabounidis, Vlad Sobal, Yangruibo Ding, Baishakhi Ray, Jonathan Buck, Guy Van den Broeck 等 12 位作者
Coding agents are typically provided end-to-end behavioral specifications that the code they produce should satisfy, such as tests and input--output examples. How many such specifications must a developer provide, and which ones, for the agent to produce the developer's intended program? We argue that this cannot be determined a priori, for it depends on the structure of the implementation, which is left to the agent to decide. An agent first commits to a structure---a decomposition into data types, modules, and functions---which fixes a class of candidate programs; behavioral specifications then select a program within that class. A set of tests that suffices for one structure may leave another badly underdetermined, requiring different edge cases or exponentially more examples. Since the structure is chosen by the agent, developers cannot ensure their specification is complete without reverse-engineering it from thousands of lines of generated code. To resolve this, we introduce Skeleton Coding, in which developers jointly specify the structure---in high-level, code-like format---alongside behavioral specifications. Theoretically, we model AI-assisted coding as a control problem, and show that skeletons can reduce exponentially the number of tests developers must supply to steer an agent toward an acceptable program. Empirically, across four models and three benchmarks, the best-performing skeleton specifications reduce mean test error by 48.4% relative to the best end-to-end behavioral specifications. Even when the agent can query a complete reference implementation, adding a skeleton reduces mean test error relatively by up to 27.0%. Skeletons also narrow the average gap between small and large models by about 52.4% while lowering implementation cost by 15--26x, and reduce code erosion by 24.7% on average.
We present EasyProver, a Lean proving system built from recursive agents with isolated contexts, asynchronous search, parallel tactic probing, and an elastic elaboration service. Using an off-the-shelf frontier model with no additional training, it proves 669/672 PutnamBench problems and all six IMO 2026 formalizations and resolves 597/600 NTP4VC verification conditions and all 189 VERINA tasks by proof or refutation. Because these benchmarks are close to saturation, we also introduce Aegis, which converts competitive programming problems into much harder verification tasks over real implementations. The pipeline writes safe Rust solutions, translates them into Lean with Aeneas, generates specifications, and labels each frozen statement with a kernel-checked proof or refutation. It produces 2,045 differentially validated Rust-Lean program pairs and 11,533 candidate specifications. Unlike hand-written benchmarks, these tasks carry implementation-level obligations from fixed-width arithmetic, explicit failure, generated loops, casts, and functionalized mutation. On a stratified sample of 2,000 statements, EasyProver proves 1,054 (52.7%) within a 30-minute budget per statement. The same tasks can also be used to benchmark other provers and as rewards for reinforcement learning.
Generative AI (GenAI) makes open source software (OSS) contributions cheaper to produce, and it makes them say less about the people who produced them. In OSS mentoring, that matters: mentors judge what newcomers understand from their work, and newcomers learn from mentors' feedback. We studied how GenAI is changing mentoring in Google Summer of Code (GSoC), a program in which today's mentees can become tomorrow's mentors. We screened every thread on GSoC's restricted mentors' mailing list from November 2022 through March 2026 and interviewed 15 mentors and organization administrators. GenAI helped with onboarding, coordination, and first-pass review, but proposals, entry contributions, and mentees' answers became weaker evidence of what newcomers could do. Mentors made up the difference by checking, and that checking drew on the same time and goodwill the program needs to renew itself; one organization had withdrawn from GSoC. Responses ranged from banning GenAI to accepting it, and the boundary an organization drew determined who did the checking. We propose AIM, a decision framework that links acceptable GenAI use, integration into the mentoring pipeline, and mentoring capacity, to help organizations choose responses they can sustain.
Large Language Models (LLMs) have demonstrated remarkable potential in code generation. The integration of Chain of Thought (CoT) reasoning can further boost their performance. However, current CoT methods often require manual writing or LLMs with over 100 billion parameters to generate, impeding their applicability in resource-constrained scenarios. In this study, we investigate lightweight Language Models (lLMs), which are defined to have fewer than 10 billion parameters. Empirically, we find that most lLMs cannot generate high-quality CoTs when prompted by the few-shot method, but can take advantage of high-quality CoTs generated elsewhere to improve their performance in code generation. Based on these findings, we design a novel approach COTTON which can leverage lLMs to automatically generate CoTs for code generation. We synthesize new datasets and conduct extensive experiments on various benchmarks. The results show that the CoTs generated by COTTON outperform the baselines in terms of automated and human evaluation metrics. In particular, the CoTs generated by COTTON boost various lLMs to achieve higher performance gains than those generated by LLMs such as ChatGLM (130B), and are competitive with those generated by gpt-3.5-turbo (175B). Our study also showcases the potential of lLMs in software engineering applications.
A dynamic invariant detector proposes invariants, such as method pre- and postconditions, and rules out those that do not hold at run time. Traditional detectors propose invariants from a fixed library of templates, which limits their expressiveness. A large language model (LLM) can propose invariants instead of, or in addition to, template-derived ones. With the right prompt and context choices, we found that this idea can both improve the invariants' expressiveness and reduce the cost of inferring the invariants. Greater expressiveness brings complications: an LLM-proposed invariant can execute arbitrary code during invariant evaluation, causing side effects or infinite loops. Our novel instrumentation strategies mitigate these problems. We wrap each candidate and evaluate only one invariant per thread at a time, so nested invariant evaluations cannot trigger one another indefinitely. To recover from non-nested hangs, we monitor each run for stalls and retry it under a growing time budget. Because we cannot prove a candidate side-effect-free in advance, we use delta debugging over the invariants evaluated during a failing test to isolate and disable a side-effecting candidate without discarding the rest of the run. We compared our implementation, Oca, to the popular template-based Daikon invariant detector on real open-source Java projects. Oca produces more expressive invariants: on average, 28% of its held invariants call a project-specific method, which Daikon's templates cannot express. Its invariants are more often valid, with a false-positive rate of 18% in a manual inspection, vs. 76% for Daikon, and more useful: as properties for property-based testing, they exposed 9 of 25 real bugs, compared with 4 for Daikon. Oca also costs less, using 566 minutes and 27 GB of disk, compared with 2,817 minutes and 827 GB for Daikon.
Evaluations of world models commonly score a predicted future state by its overlap with the ground truth. Yet in interactive worlds, this rewards copying: a model that merely preserves persistent state appears accurate without predicting any consequence of an action. We introduce ScratchWorld, an execution-grounded diagnostic benchmark that treats Scratch projects as executable worlds and derives every label from replay on a pinned Scratch VM, including next-state transitions, hidden variables, causal traces, and counterfactual outcomes. Its primary metric, value-aware changed-field F1 ($F_1^{\mathrm{VA}}$), credits a prediction only when it names a changed field and computes its executed value. Model-free anchors make the confound explicit: a no-change prior attains 91.4% full-state field accuracy on synthetic long-horizon items and 0% changed-field credit on non-empty one-step items, whereas VM replay attains 100% wherever the corresponding metric is defined. Across seven frontier LLMs prompted as world models, state-only $F_1^{\mathrm{VA}}$ reaches at most 30.0% on controlled synthetic mechanisms and declines steeply as projects approach real-world complexity. Revealing the full program raises every model's value-aware score by 22.3-51.7 points and substantially reorders the models, yet the best reaches only 55.4%, an execution gap that persists even with the rules visible. Auxiliary diagnostics complete the error profile: causal-event accuracy remains within 2.5 points of the 50% no-event baseline, with no correct exact chain; counterfactual prediction trades changed-outcome accuracy against invariance; hidden-state accuracy degrades roughly twice as fast as full-state overlap over long horizons. In a word, ScratchWorld separates what a model knows about the rules from what it can actually execute.
Souhardya Saha Dip, Moshood A. Fakorede, Krishna Upadhyay, Umar Farooq
Post-quantum cryptography (PQC) standards establish algorithms for quantum-safe systems, but empirical evidence about how software projects implement and maintain PQC support remains limited. We investigate this implementation and maintenance work by mining 114,361 commits and 5,713 issues across 179 open-source repositories. We use human annotation and validated LLM classification to identify PQC algorithms in commits, classify their adoption status and deployment mode, and identify technical challenges in issue discussions. The commit classification for PQC algorithms reveals that 90% of commits concern maintenance of existing support. From the issue discussions, we construct an 11-category taxonomy of recurring technical challenges. Build, packaging, and dependency failures, along with functional defects, account for approximately half of categorized issues. To examine how PQC support evolves, we use time-series analysis and first implementations of individual algorithms within repositories around NIST standardization milestones. After the 2024 finalization, standardized algorithm names appeared in more than 80% of the PQC development changes analyzed, although earlier names such as Kyber continued to appear. Development activity increased over the observation period. Our milestone analysis, however, did not find a significant increase in first implementations. Migration plans should therefore account for sustained maintenance, prioritize build reliability and integration testing, and track both standardized and predecessor algorithm names. We make the dataset, taxonomy, annotations, and analysis pipeline publicly available.
Changhao Li, Bradley A. Chase, Veena Vijayakumar, Mathys Rennela, Alessandro Cosentino, Naixu Guo, Sebastian Hassinger, Farrokh Labib 等 14 位作者
Quantum open-source software provides essential infrastructure for quantum science, and artificial intelligence (AI) is reshaping its capabilities and transforming its development. However, verifying quantum software at scale remains difficult, and AI-generated contributions can increase the workload of maintainers. In this Perspective, we examine verification as a central requirement for reliable AI-assisted quantum software development and research. We analyze public repositories, records covering 110 repositories from five annual quantum open-source hackathons that we organize, and responses from more than 1600 participants in a community survey. Our analysis connects limitations in automated verification with the review burden on quantum software maintainers. We propose verification practices and contribution standards that require contributors to support their changes with reproducible evidence for maintainer review. We also examine how quantum open-source software can provide infrastructure for AI-driven research whose results the community can independently verify and build upon. We maintain an open-source companion repository to support reproducibility and help projects adopt and refine these proposals. Together, these contributions help the quantum open-source software community translate advances in AI into reliable quantum research through reproducible verification.
Asher Tempest Arya, Rajan Maghera, Nicholas Wu Jiang, Federico Mora, Aren A. Babikian, Marsha Chechik
Self-adaptive systems are designed to accommodate variants of their environment, which form a family of instances whose uncertainty is captured by probabilistic models, and must meet probabilistic requirements on each instance. Within such a family, changes in size and structure complicate the fundamental adaptation decision: whether to retain, reuse, or construct a new controller, since controllers are often synthesized for and specific to an instance. Large language models (LLMs) have recently been used to write reusable controllers, but existing approaches largely target deterministic domains, and probabilistic model checking of their output remains limited to preliminary single-instance studies. In this paper, we propose a verifier-guided framework for controller reuse with explicit support for families of environments. The framework introduces a novel combination of LLM-written controllers with probabilistic model checking to provide per-instance guarantees for probabilistic requirements. We evaluate our approach using three LLMs on three benchmarks, assessing its effectiveness and cost with respect to baseline implementations, the distribution of adaptation decisions, and the impact of individual framework components. On grid worlds, LLM-written rule lists are verified on about 90% of held-out instances, against at most 77% for the baselines, and verifying a stored rule list is 15 to 103 times faster than per-instance synthesis. Reuse serves most episodes at a small fraction of the cost of generation or per-instance synthesis, and model-checking feedback increases verified coverage over independent restarts.
Jenny T. Liang, Mihika Bairathi, Wayne Chi, Ameet Talwalkar, Nishant Subramani, Valerie Chen
Imperfections in AI-generated code require that software developers modify generations through manual code edits. However, little is known about how developers edit AI-generated code, and to what extent LLMs can infer implicit developer intent from this data, given the lack of high-quality, realistic code editing data. This information is essential for designing and building usable AI-developer interactions. To address this gap, we introduce DECODE (Developer Edits of Code Dataset), a dataset of 53.6K real-world in-IDE code edits of AI-generated code sourced from 1K+ developers. First, we use DECODE to analyze when, why, and how AI-generated code is edited. Developers invest little effort in editing code and abandon AI-generated code quickly, thereby motivating the need for LLMs to understand developer intents through their edits. Therefore, we next use DECODE to understand to what extent LLMs could infer developer edits. Predicting code edits is a challenging task for both frontier and fine-tuned models. Frontier models generate edits that best reflect developer intents, but are prone to over-editing and reverting developer-written edits. Fine-tuning LLMs can improve their ability to predict edits that are more similar to developers' as well as better reflect developer intent, but do not fully bridge the gap to frontier models. Together, our results demonstrate the unique insights on developer and LLM behavior that can be obtained through studying realistic code edit data to promote more better developer-AI interactions.
Maintaining or improving the performance of Deep Neural Networks (DNNs) through fine-tuning requires labeling newly collected inputs, a process that is often costly and time-consuming. To alleviate this problem, input selection approaches have been developed in recent years to identify small, yet highly informative subsets for labeling. Diversity-based selection is one of the most effective approaches for this purpose. However, they are often computationally intensive and lack scalability for large input sets, limiting their practical applicability. To address this challenge, we introduce Concept-Based Diversity (CBD), a novel and highly efficient diversity metric for image inputs that leverages Vision-Language Models (VLMs). Our results show that CBD exhibits a strong correlation with Geometric Diversity (GD), an established diversity metric, while requiring only a fraction of its computation time. Building on this finding, we propose a hybrid input selection approach that combines CBD with Margin, a simple uncertainty metric. We conduct a comprehensive evaluation across a diverse set of DNN models, input sets, selection budgets, and six most effective state-of-the-art selection baselines. The results demonstrate that the CBD-based selection consistently outperforms all baselines at guiding input selection to improve the DNN model. Furthermore, the CBD-based selection approach remains highly efficient, requiring selection times close to those of simple uncertainty-based methods such as Margin, even on larger input sets like ImageNet. These results confirm not only the effectiveness and computational advantage of the CBD-based approach, particularly compared to hybrid baselines, but also its scalability in repetitive and extensive input selection scenarios.
Fast-paced AI-assisted software development risks reducing developers' understanding of the systems they build. This growing gap between what the system does and what developers understand about it, described as ``cognitive debt'', is not new, but AI has exacerbated it. This has led to a recent surge in blog posts and online discussions, in which practitioners recognize and write about cognitive debt, and share their experiences well before academic studies can capture them. To understand their concerns and inform future research and practice, we conduct a grey literature review of 76 sources, focusing on how cognitive debt is conceptualized and manifests, its contributing factors, its consequences, and the practices to prevent, detect, and mitigate it. We find that cognitive debt can accumulate unnoticed behind working code and passing tests, particularly when developers accept AI-generated solutions without building understanding. Practitioners report consequences such as harder maintenance and debugging, fragmented knowledge in the team, and reduced productivity. They propose practices such as documenting design rationale, review-sized changes, comprehension gates, and knowledge-sharing rituals, along with signals like reliability metrics to detect it. Our findings give software teams a structured view of emerging practical approaches in AI-assisted development and stress the need to preserve human understanding alongside faster code production.
Shenglong Yao, Zirui Liu, Xiangchi Yuan, Wenke Lee
Recent benchmarks show that LLMs achieve strong performance in both function-level binary understanding and end-to-end binary reverse engineering tasks, but whether LLMs can reason over interprocedural semantics in compiled binaries remains unclear. Function-level benchmarks offer limited visibility into cross-function relationships, while whole-program benchmarks typically evaluate final outcomes of tasks that may be solvable from local clues alone, making it difficult to disentangle interprocedural reasoning from exploration, tool use, execution, and retrieval. To address this limitation, we introduce ICALL-BENCH, a benchmark that directly evaluates whether LLMs and agents can understand and reason over interprocedural semantics in binary programs, using indirect-call target recovery as a measurable probe. ICALL-BENCH covers 50 complete real-world C/C++ programs spanning different complexity levels and multiple optimization configurations, with an average of 74,746 lines of code and 941 functions per program. We further introduce three analysis regimes spanning model-only reasoning, static tool-augmented agents, and Internet-enabled open-world agents to examine how intrinsic reasoning, active information acquisition, and external knowledge each contribute to interprocedural binary analysis. We evaluate five recent LLMs and their corresponding agents. Our results reveal a wide separation across current models and agents, with program-level verifiable success rates ranging from 23% to 81%. They also show that greater information access does not consistently lead to better results, with particularly limited gains on the hardest cases, where deeper interprocedural reasoning remains the primary bottleneck. Uneven performance across models and frequent verifiable failures even under an optimistic evaluation indicate that robust binary interprocedural reasoning remains an open challenge.
As AI coding agents take on increasingly complex, long-horizon software engineering tasks, existing benchmarks are rapidly saturating and their evaluation quality has come under serious scrutiny: a recent audit found that nearly 60% of unsolved SWE-bench Verified instances contain flawed tests -- either overly narrow tests that reject correct solutions or overly broad tests that check unstated requirements -- and that frontier models can verbatim reproduce gold patches from training data. Code refactoring, which requires coordinated, behavior-preserving changes across many files, offers a substantially harder and more realistic test of agent capability, yet remains underserved by current benchmarks. We introduce SWE-Bench ProMax, an expert-curated, multilingual code refactoring benchmark of 170 instances drawn from real commits across seven programming languages (Python, Java, TypeScript, Go, C, C++, and Rust). Every instance undergoes rigorous, multi-stage curation that directly addresses the quality problems identified in prior benchmarks: issue descriptions are rewritten from scratch to provide precise, unambiguous specifications, and test suites are manually reviewed to remove overly narrow and overly broad tests. Tasks with insufficient complexity or limited cross-file scope are filtered out, yielding a benchmark of challenging, large-scale refactoring tasks that average 11.5 modified files and 261.6 lines of code per instance, substantially exceeding the scale of existing benchmarks. Experiments with frontier models under two agent scaffolds show that the best model achieves only 41.2% resolve rate, confirming that SWE-Bench ProMax presents a meaningful and unsaturated challenge for current AI coding agents. Our benchmark is available at https://huggingface.co/datasets/swe-bench-promax/SWE-Bench-ProMax.
Abhinav Goel, Agostino Capponi, Alfio Gliozzo, Chaitya Shah
AI agents can now write software from a plain-English description, but code that runs is not the same as code that does what was asked. This gap matters most for smart contracts, the programs behind Web3 applications that move money on a blockchain and cannot be fixed once deployed. We ask how to check whether an AI-written contract behaves as it should, and where the requested behavior gets lost. We design an evaluation built on execution rather than opinion. We replay the same transactions on an AI-written contract and on the original it should reproduce and measure how often they agree. We test every grading tool on planted mistakes, compare against an answer key written without the description, and audit our own scorer until known wrong answers cannot pass. Common methods, such as AI judges and standard analysis tools, miss obvious mistakes, and scoring choices matter too. Checking only the functions every system includes hid about half the difference between pipelines. We then propose an elicitation-first architecture. Before writing code, the agent asks the client a short checklist of questions, checks the contract against an independently written executable answer key, and returns disagreements to the client instead of silently rewriting code. Asking first matters far more than reviewing or patching afterwards, across three AI models, while automated security fixes often break requested behavior. A few well-chosen questions recover much of what a short description omits, and their value can be estimated without the original contract. To our knowledge, this is the first study in smart contract synthesis to gather missing requirements before generation and check the result against an independently written executable model. The lessons extend to any setting where AI agents turn requests into programs. We release our tools, 600 tasks and all results.
Software vulnerability detection remains difficult because existing approaches rely on predefined analysis rules, learn vulnerability patterns implicitly from code and labels, or use large language models (LLMs) to make the final security judgment. It remains unclear whether structured descriptions of security-relevant program behavior can improve supervised detection. We present VulnLens, an explanation-augmented framework for binary C/C++ vulnerability detection that separates program-behavior analysis from the final decision. A label-blind LLM-based Explainer examines each function from seven fixed perspectives: data flow, memory access, object lifetime, arithmetic and type behavior, control flow, shared state, and external-resource access. Without receiving vulnerability labels or CWE information, it generates structured explanations. A supervised Detector combines source code and explanations through sequential cross-attention. We evaluate VulnLens on a PrimeVul-derived dataset covering ten CWE categories and test transfer to SecVulEval without retraining. On the 526-function PrimeVul test set, VulnLens achieves 75.5% accuracy and 77.2% F1, exceeding DeepSeek V4 Pro by 4.2 percentage points in accuracy and Gemini 2.5 Pro by 2.2 points in F1. Matched explanations improve F1 by 7.3 points over a separately trained code-only detector. On SecVulEval, VulnLens achieves 66.4% accuracy and 65.8% F1, outperforming DeepSeek V4 Pro by 7.6 points in accuracy and 20.0 points in F1, and Gemini 2.5 Pro by 16.6 points in accuracy and 7.8 points in F1.
Liliana Hotsko, Yinxi Li, Yuntian Deng, Pengyu Nie
Code language models need repository-level context to resolve imports, APIs, and project conventions. Existing methods inject this knowledge as long inputs (retrieved through RAG or dependency analysis) or through per-repository fine-tuning and LoRA---costly at repository scale and brittle to evolving codebases. We introduce Code2LoRA, a hypernetwork framework that generates repository-specific LoRA adapters, effectively injecting repository knowledge with zero inference-time token overhead. Code2LoRA supports two usage scenarios: Code2LoRA-Static converts a single repository snapshot into an adapter, suitable for comprehension of stable codebases; while Code2LoRA-Evo maintains an adapter backed by a GRU hidden state updated per code diff, suitable for active development of evolving codebases. To evaluate Code2LoRA against parameter-efficient fine-tuning baselines, we build RepoPeftBench, a benchmark of 604 Python repositories with two tracks: a static track with 40K training and 12K test assertion-completion tasks, and an evolution track with 215K commit-derived training and 87K commit-derived test tasks. On the static track, Code2LoRA-Static achieves 63.8% cross-repo and 66.2% in-repo exact match, matching the per-repository LoRA upper bound; on the evolution track, Code2LoRA-Evo achieves 60.3% cross-repo exact match (+5.2 pp over a single shared LoRA). Code2LoRA's code can be found at https://anonymous.4open.science/r/code2lora-6857; the model checkpoints and RepoPeftBench datasets can be found at https://huggingface.co/code2lora.
An EVM-Cosmos Bridge for Atomic, Semantic-Preserving Transfer of Environmental Tokens
Sebastian Pandolfi, Guzmán Llambías, Laura González
AI 速览：提出一种在 EVM 与 Cosmos 之间转移环境代币的跨链桥，用确定性指标身份防止语义混淆，并扩展 HTLC 实现单所有者原子转移。
Transferring semantically typed tokens between heterogeneous blockchains (e.g. EVM and Cosmos ecosystems) raises two problems: semantic mixing, the loss of a token's domainspecific meaning in transit (e.g., CO2 removals issued under different methodologies), and atomicity despite divergent consensus and timing assumptions. We propose a bridge that addresses both. A deterministic indicator identity derived from cryptographic commitments, mirrored in class registries on both chains and checked at mint time, preventing semantic mixing by construction. For atomicity, we extend hash time-locked contracts (HTLCs) from two-party exchange to single-owner transfer with an automatic callback that burns the source escrow after a destination claim, and a protocol-minimum bounty that makes the fallback burn incentive-compatible even against adversarial senders. We machine-check the protocol's invariants with TLC (supply conservation fails without the rational-monitor assumption and holds with it) and validate a bidirectional prototype on Polygon and Neutron mainnets over Axelar GMP: fully automatic transfers with a median end-to-end latency of 97 s over 50 Polygon to Neutron transfers across 7 days (all completed by the on-chain automatic callback) and 93 s in the reverse direction, and semantic-validation overhead below 0.4% of messaging cost. We also document the integration failure modes of Cosmos originated messaging, whose dominant operational risk proves to be relay eligibility rather than latency.
As Large Language Models (LLMs) are increasingly integrated into software development workflows, they also become prime targets for adversarial attacks. Among these, backdoor attacks are a significant threat, allowing attackers to manipulate model outputs through hidden triggers embedded in training data. Detecting such backdoors remains a challenge, and one promising approach is the use of Spectral Signature defense methods that identify poisoned data by analyzing feature representations through eigenvectors. While some prior works have explored Spectral Signatures for backdoor detection in neural networks, recent studies suggest that these methods may not be optimally effective for code models. In this paper, we revisit the applicability of Spectral Signature-based defenses in the context of backdoor attacks on code models. We systematically evaluate their effectiveness under various attack scenarios and defense configurations, analyzing their strengths and limitations. We found that the widely used setting of Spectral Signature in code backdoor detection is often suboptimal. Hence, we explored the impact of different settings of the key factors. We discovered a new proxy metric that can more accurately estimate the actual performance of Spectral Signature without model retraining after the defense. Our findings contribute to a deeper understanding of the configurations of Spectral Signature defenses in securing code models and offer insights into potential improvements for future research.
Enterprises increasingly mandate the use of AI coding tools and announce ambitious productivity targets, yet there is little longitudinal evidence on whether such productivity targets have been reached, how, and at what cost. We present a quantitative case study of a documented enterprise "2x" mandate: in June 2025, the CTO of a mid-sized, AI-forward software company set the goal of doubling merged pull requests (PRs) per engineer with AI. Using a panel of 802 developers and 196,212 PRs (January 2024 - April 2026), we estimate the effect of AI adoption by comparing each developer's PR throughput after they started using AI tools with their own throughput before. We find that the company eventually reached the target, but only in the tenth month and not in the way the announcement implied: PR throughput barely moved at the mandate date, rose by about 12% when each developer started using AI, and then kept growing as developers accumulated use. The target was not met merely by working longer or splitting work: extra working days account for at most one-fifth of the throughput increase, and the added PRs became larger and skewed toward features and bug fixes rather than routine changes. The gain is similar across seniority levels and larger in repositories created since 2022. The cost of AI adoption fell on code review, where load per reviewer doubled and automated review overtook human review in coverage within months of the mandate. We interpret these results as strong evidence that the adoption and sustained usage of AI raises PR throughput in an enterprise setting, with two caveats: throughput is only a proxy for productivity, and the exact size of the effect remains uncertain due to a lack of truly randomized assignment in our setting. Our results have implications for organizations under AI transformations and researchers studying the impact of AI on software development productivity.