公开论文雷达

We introduce TCS-Bench, a benchmark for evaluating Large Language Models (LLMs) on research-level Theoretical Computer Science (TCS) proof generation. TCS-Bench consists of theorem-proving tasks from papers published at top theoretical computer science venues (STOC, FOCS, and SODA). Each task provides the necessary context to derive a self-contained proof for a target result. We evaluate state-of-the-art models on this benchmark. We verify the correctness of generated proofs via a verification agent, and further benchmark the verifier against human-expert proof judgements on a set of target statements and generated proofs pairs. Our reference verifier achieves over 90% accuracy on the expert labeled set.

Verified code generation asks a large language model (LLM) to generate both an executable program and a machine-checkable proof that the program meets a formal specification, promising software that is correct by construction. The de facto workflow decouples the two halves of the problem: first synthesize a program, then attempt to prove it correct. We observe that this sequential pipeline can be both ineffective and inefficient in practice. A program generated without anticipating its proof can be subtly incorrect or structurally difficult to verify, forcing the LLM into brittle repair loops that alternate between patching the code and patching the proof. Inspired by Dijkstra's view that a program and its correctness argument should be developed hand in hand, we propose $P^3$, an LLM-based agentic workflow that first derives a unified program-and-proof plan from the specification, then elaborates the implementation and proof scaffold under this shared plan. To evaluate verified code generation in realistic settings, we further introduce Lean4Commit0, a repository-derived, library-level benchmark built by extracting core APIs from real-world software repositories and translating their requirements, including relational specifications across APIs, into Lean tasks. Using four frontier LLM backends, we evaluate $P^3$ on Verina, AlgoVeri, and our Lean4Commit0 benchmark, where it achieves the highest solve rate in every benchmark--model setting. Compared with the stronger baseline, it improves solve rates by 4.6--11.2 percentage points and reduces per-task API cost by up to roughly 40\% and wall-clock time by up to roughly 37\% on the difficult subset of each benchmark. A targeted ablation further shows gains of 3.3--8.3 points over implementation-only planning, isolating the benefit of planning the program and proof jointly.

Compiler optimizations are essential for achieving high performance in modern software. However, recent studies highlight the persistence of performance bugs, i.e., subtle defects where the compiler generates functionally correct but computationally inefficient code, leading to significant performance degradation. Existing detection and testing methods typically employ a bottom-up approach, focusing on specific low-level code properties and remaining confined to known optimization rules. Consequently, they struggle to quantify the holistic impact of identified issues and often overlook critical microarchitectural inefficiencies. We observe a key indicator of untapped potential: different compilers often produce binaries with significant performance differences for identical source code. However, the root causes of these discrepancies remain largely unexplored and difficult to pinpoint using current techniques. To bridge this gap, we introduce a top-down differential analysis methodology. This approach calibrates compiler optimization differences with fine-grained, hierarchical microarchitectural metrics, offering a comprehensive view of runtime behavior. Using a sampling-based approach, this method efficiently pinpoints the critical code snippets responsible for performance differences, enabling targeted root cause analysis. Our empirical evaluation uncovers substantial and often surprising performance differences between binaries generated by GCC and Clang. A categorization of root causes reveals systemic challenges in compiler optimizations. To quantitatively validate our findings and demonstrate practical impact, we developed a binary patching framework that fixes identified performance issues by transplanting superior code sequences from competing compilers. This work provides a novel lens for understanding and analyzing optimization defects.

LLM-based code review agents promise scalable, always-on review, yet current systems suffer from two intertwined weaknesses: (1) non-determinism--unbounded tool use makes review outcomes unstable, and (2) context locality--the reviewer's access remains bounded to the diff, capping discoverable issue depth. Both give rise to three challenges: misaligned context retrieval, a coherence-efficiency trade-off in multi-file pull requests, and hallucinated comments that erode trust. To address these, we introduce OpenCodeReview, built on deterministic engineering for uncertain agents: rather than granting maximal freedom, we inject determinism at three deliberate pipeline points. Rule-Guided Dispatch uses a multi-layer rule system to deterministically select files and review criteria, eliminating variability of agent-driven triage. Grounded File Review replaces free-form exploration with a curated tool set exposed through a ReAct loop, while file-level parallel SubAgents balance context coherence against efficiency and recover cross-file dependencies on demand. Independent Reflection introduces a falsification-first filter under an asymmetric information boundary--the reflector sees only the diff, not the agent's tool-augmented exploration--removing hallucinated comments without self-reinforcing bias, improving precision while preserving recall. On AACR-Bench (200 real-world PRs, 10 languages, 1,505 expert-verified comments), OpenCodeReview outperforms mainstream coding agents (e.g., Claude Code and Codex) across six LLM backends, achieving up to 2.17x higher SEM-F1 (25.10% vs. 11.57%) while consuming 5-15x fewer tokens. We open-source OpenCodeReview at https://github.com/alibaba/open-code-review.

LLMs have demonstrated strong capabilities in code generation and automated program repair, but migrating an entire repository rarely produces a runnable application because long-horizon translation challenges LLM-based agents' ability to maintain repository-level migration objectives. In this work, we propose Entropy-based Code Adversarial Translation (ECAT), a multi-agent framework for automated Android-to-HarmonyOS repository migration. ECAT formulates repository migration as adversarial entropy minimization through a generator-discriminator architecture. The discriminator measures migration quality using a unified metric called Code Entropy and produces text gradients that specify both file-level generation directives and the skills needed to execute them. Guided by these optimization signals, the generator iteratively updates the repository, and each update is accepted only if it reduces Code Entropy. Repeated generator--discriminator interactions progressively drive the migration from an initial template toward a functionally complete HarmonyOS repository. Successful low-entropy trajectories are further distilled into a self-evolving memory tree, enabling transferable migration knowledge across repositories. We also introduce A2H-RepoBench, the first real-world benchmark for Android-to-HarmonyOS repository migration, covering applications from tens of thousands to hundreds of thousands of lines of code. Evaluated by node alignment and an agent-based functional judge, ECAT achieves 74.7% overall migration quality and consistently outperforms existing agent-based methods across repositories of different scales.

Code generated by modern language models often reads naturally. Yet, it also often fails to implement what was asked. This should be no surprise, as research shows the models' own confidence signals are poorly calibrated with actual correctness. A promising way to assess correctness looks inside the model: by contrasting the hidden states of correct and incorrect programs, recent work captured an internal signal of code correctness that is able to judge candidate solutions better than the model's token-level or stated confidence, with no test execution. However, this signal was captured under one particular way, leaving open an important question: whether it reflects a robust property of the model or an artifact of that choice. We study this question systematically, varying how the signal is extracted from the model internals. Besides this, we also ask if the signal's quality is limited by the data used to extract it, by constructing program pairs that differ only in the fault that makes them incorrect. Our results show that no single configuration is best, and that isolating the fault does not help.

Code smells significantly threaten software maintainability by accumulating technical debt, yet developers often lack the resources to manually address these flaws under tight release schedules. While static analysis tools like SonarQube provide precise detection, they function largely as passive alert systems, leaving the burden of refactoring on developers. To bridge this gap, we present a novel cleaning tool, namely SmellCC, a Visual Studio Code extension that augments SonarQube with an LLM-based pipeline to automatically detect and refactor Python code smells. By employing Chain-of-Thought (CoT) and few-shot learning, SmellCC provides in-place, one-click remediation for the top-10 most frequent smells, effectively preventing the accumulation of technical debt during development. Our quantitative evaluation demonstrates that our SmellCC is promising in helping developers effectively eliminate code smells (96.8\% cleaning rate) with high accuracy (i.e., 91.3\%), ensuring that the refactored code remains syntactically correct and behavior-preserving, thereby significantly improving long-term software maintainability.

Code performance optimization is a vital aspect of modern software development, as it enables faster response times and reduced resource usage. These optimizations require a deep understanding of low-level hardware details and the intricacies of parallel processing, making them challenging even for experienced developers. With the advent of Large Language Models (LLMs), which are increasingly capable of generating and understanding code, there is growing interest in incorporating these models into automated code optimization processes. Traditionally, this automation involves transcribing the source code into a domain-specific representation that can be auto-tuned using grid search or machine learning algorithms, while adhering to strict rules and a limited set of feasible transformations to ensure verifiability. LLMs incorporate high-level code semantics and can thus perform transformations that go beyond verifiable automated optimizations. This paper investigates whether the traditional abstractions used in automated code optimization improve the performance and correctness of LLM-guided optimizations of parallel HPC applications. We evaluate this using the PolyBench benchmark suite and demonstrate that, in our evaluated setting, LLMs provided with specific optimization goals achieve better measured performance and validity rates when generating C code compared to creating computation pipelines and optimization schedules with established frameworks, suggesting that future development should explore alternative approaches for verifiable LLM-guided code optimization.

Recent advances have highlighted the potential of machine learning, particularly Large Language Models (LLMs), for analyzing and optimizing programs. We present the first application of program embeddings from LLMCompiler---an LLM massively pretrained on intermediate representation (IR) code---to representative program analysis and optimization tasks. We generate program embeddings directly from source and IR code using a simple approach: split programs into chunks, independently embed each chunk with pretrained LLMs, and then aggregate the chunk embeddings into a single program embedding. Our experiments show that combining source and IR code embeddings achieves an error rate of 1.54\% in algorithm classification, a 12\% improvement over the current state-of-the-art, and a competitive accuracy on heterogeneous device mapping. These findings suggest that training a performance-aware LLM for embedding IR code might yield state-of-the-art results in code optimization tasks.

User Interface (UI) programming is challenging due to the complex abstraction gap between code and graphical software representations. To bridge this gap, UI programming tools often rely on screen retrieval and clustering, which require accurate similarity measures based on overlapping features. However, computing feature-oriented similarity is difficult because screens with similar functionality often exhibit design variations. To address this, we propose FRAME (ReinForced UseR InterfAce Screen EMbedding with Graphical Structural ComprEhension), a multi-modal, neuro-symbolic embedding technique. FRAME constructs symbolic, graph-based representations of UI components to encode salient relationships and capture feature patterns across different screens. It leverages large vision-language models for visual and lexical encoding, alongside a novel UI-specific computational geometry algorithm that enables weighted embedding propagation. Across three benchmarks, FRAME outperforms strong baselines by up to 13% MRR in search and 7.6 percentage points in clustering accuracy. A comprehensive ablation study further confirms the benefit of each component, demonstrating FRAME's potential for enhancing automated UI design and testing tools.

AI agents are becoming shared infrastructure, yet durable memory is commonly assembled from separate retrieval, governance, and operational components. We present SuperLocalMemory 4.0, a governed, local-first memory operating system for AI agents. The system combines dense semantic, BM25 lexical, temporal, Hopfield-associative, and spreading-activation retrieval through reciprocal-rank fusion; a governed learning and behaviour layer; bi-temporal recall; multi-scope personal, shared, and global memory; role-based access control; GDPR-oriented export and verified erasure; audit trails; and a deployment-context EU AI Act checklist. V4 introduces a reliability spine for its primary write path: generation-fenced admission, a policy registry, verifiable memory transactions with per-projection apply, verify, compensate, and erase owners, and hash-checkable completion manifests. The runtime is available through CLI, MCP, an HTTP daemon, a dashboard, editor integration, and framework adapters, and supports fully local, local-with-on-device-model, and provider-assisted modes. We evaluate eleven fault-injection and mechanism scenarios, each repeated 200 times. The released evidence bundle reports 2,200 of 2,200 deterministic repetitions upholding their scoped component properties. The governed write envelope measured 3.522 ms at p50 and 5.297 ms at p99, versus 1.835 ms and 2.569 ms for the ungoverned baseline, corresponding to in-process control-plane overheads of 1.687 ms at p50 and 2.728 ms at p99. These are scoped component and mechanism measurements, not an end-to-end multi-process or external retrieval-accuracy benchmark. The paper consolidates prior SuperLocalMemory work on privacy-preserving multi-agent memory, information-geometric retrieval, and the V3.3 Living Brain lifecycle.

Long-term memory enables AI agents to maintain continuity across sessions, personalize behavior, and evolve through accumulated experience. Yet memory evolution is not simply a process of storing more information: new experiences may reinforce, revise, or interfere with existing memory states. Existing systems mainly emphasize memory construction and relevance-based retrieval, but several memories may remain simultaneously relevant while differing in state, temporal validity, or authority. We introduce Controlled Memory Interference (CMI), a controlled diagnostic and data-generation framework for studying how agent memory evolves under different memory relationships. Across controlled memory evolution, benign accumulation has limited effects, whereas relationship-specific interference sharply suppresses update plasticity with little stability gain, either by blocking target-memory exposure or by disrupting its downstream use. Lexical and Dense retrieval exhibit distinct interference pathways, while poisoning is more sensitive to update-authority cues than to recency alone. Beyond diagnosis, CMI provides targeted examples for interference-aware memory learning, improving the distinction between valid updates and interference-inducing memories while preserving performance on original memory tasks. These findings show that memory evolution is shaped not only by memory scale, but also by interactions among accumulated experiences. More broadly, memory interference emerges as an important factor for reliable continual agent memory systems.