Integration-by-parts reductions of Feynman integrals pose a frequent bottle-neck in state-of-the-art calculations in theoretical particle and gravitational-wave physics, and rely on heuristic approaches for selecting integration-by-parts identities, whose quality heavily influences the performance. In this paper, we investigate the use of machine-learning techniques to find improved heuristics. We use funsearch, a genetic programming variant based on code generation by a Large Language Model, in order to explore possible approaches, then use strongly typed genetic programming to zero in on useful solutions. Both approaches manage to re-discover the state-of-the-art heuristics recently incorporated into integration-by-parts solvers, and in one example find a small advance on this state of the art.
Existing methods fail to effectively steer Large Language Models (LLMs) between textual reasoning and code generation, leaving symbolic computing capabilities underutilized. We introduce CodeSteer, an effective method for guiding LLM code/text generation. We construct a comprehensive benchmark SymBench comprising 37 symbolic tasks with adjustable complexity and also synthesize datasets of 12k multi-turn guidance/generation trajectories and 5.5k guidance comparison pairs. We fine-tune the Llama-3-8B model with a newly designed multi-turn supervised fine-tuning (SFT) and direct preference optimization (DPO). The resulting model, CodeSteerLLM, augmented with the proposed symbolic and self-answer checkers, effectively guides the code/text generation of larger models. Augmenting GPT-4o with CodeSteer raises its average performance score from 53.3 to 86.4, even outperforming the existing best LLM OpenAI o1 (82.7), o1-preview (74.8), and DeepSeek R1 (76.8) across all 37 tasks (28 seen, 9 unseen). Trained for GPT-4o, CodeSteer demonstrates superior generalizability, providing an average 41.8 performance boost on Claude, Mistral, and GPT-3.5. CodeSteer-guided LLMs fully harness symbolic computing to maintain strong performance on highly complex tasks. Models, Datasets, and Codes are available at https://github.com/yongchao98/CodeSteer-v1.0 and https://huggingface.co/yongchao98.
In this paper, we investigate how the order in which tokens are unmasked during masked diffusion models (MDMs) inference affects generative quality. We derive an expanded evidence lower bound (ELBO) that introduces a planner, responsible for selecting which tokens to unmask at each step. Our analysis suggests that alternative unmasking strategies can improve generative performance. Based on these insights, we propose Path Planning (P2), a sampling framework that leverages pre-trained BERT or the denoiser itself to guide unmasking decisions. P2 generalizes all known MDM sampling strategies and enables significant improvements across diverse domains including language generation (in-context learning, code generation, story infilling, mathematical reasoning, reverse curse correction) and biological sequence generation (protein and RNA sequences).
Capability evaluations are required to understand and regulate AI systems that may be deployed or further developed. Therefore, it is important that evaluations provide an accurate estimation of an AI system's capabilities. However, in numerous cases, previously latent capabilities have been elicited from models, sometimes long after initial release. Accordingly, substantial efforts have been made to develop methods for eliciting latent capabilities from models. In this paper, we evaluate the effectiveness of capability elicitation techniques by intentionally training model organisms -- language models with hidden capabilities that are revealed by a password. We introduce a novel method for training model organisms, based on circuit breaking, which is more robust to elicitation techniques than standard password-locked models. We focus on elicitation techniques based on prompting and activation steering, and compare these to fine-tuning methods. Prompting techniques can elicit the actual capability of both password-locked and circuit-broken model organisms in an MCQA setting, while steering fails to do so. For a code-generation task, only fine-tuning can elicit the hidden capabilities of our novel model organism. Additionally, our results suggest that combining techniques improves elicitation. Still, if possible, fine-tuning should be the method of choice to improve the trustworthiness of capability evaluations.
Long-form generation is crucial for academic writing papers and repo-level code generation. Despite this, current models, including GPT-4o, still exhibit unsatisfactory performance. Existing methods that utilize preference learning with outcome supervision often fail to provide detailed feedback for extended contexts. This shortcoming can lead to content that does not fully satisfy query requirements, resulting in issues like length deviations, and diminished quality. In this paper, we propose enhancing long-form generation by incorporating process supervision. We employ Monte Carlo Tree Search to gather stepwise preference pairs, utilizing a global memory pool to maintain consistency. To address the issue of suboptimal candidate selection, we integrate external critiques to refine and improve the quality of the preference pairs. Finally, we apply step-level DPO using the collected stepwise preference pairs. Experimental results show that our method improves length and quality on long-form generation benchmarks, with almost lossless performance on general benchmarks across various model backbones.
This paper investigates an underexplored challenge in large language models (LLMs): the impact of KV cache compression methods on LLMs' fundamental capabilities. Although existing methods achieve impressive compression ratios on long-context benchmarks, their effects on core model capabilities remain understudied. We present a comprehensive benchmark KVFundaBench to systematically evaluate the effects of KV cache compression across diverse fundamental LLM capabilities, spanning world knowledge, commonsense reasoning, arithmetic reasoning, code generation, safety, and long-context understanding and generation.Our analysis reveals serval key findings: (1) \textit{Task-Dependent Degradation}; (2) \textit{Model-Type Robustness} (3) \textit{Prompt Length Vulnerability}; (4) \textit{Chunk-Level Superiority}; (5) \textit{Prompt-Gain Sensitivity}; (6) \textit{Long-Context Generation Sensitivity}. Based on our analysis of attention patterns and cross-task compression performance, we propose ShotKV, a novel compression approach that distinctly handles prefill and decoding phases while maintaining shot-level semantic coherence. Empirical results show that ShotKV achieves $9\%$-$18\%$ performance improvements on long-context generation tasks under aggressive compression ratios.
Foundation models (FMs), particularly large language models (LLMs), have shown significant promise in various software engineering (SE) tasks, including code generation, debugging, and requirement refinement. Despite these advances, existing evaluation frameworks are insufficient for assessing model performance in iterative, context-rich workflows characteristic of SE activities. To address this limitation, we introduce SE Arena, an interactive platform designed to evaluate FMs in SE tasks. SE Arena provides a transparent, open-source leaderboard, supports multi-round conversational workflows, and enables end-to-end model comparisons. The platform introduces novel metrics, including model consistency score that measures the consistency of model outputs through self-play matches, and conversation efficiency index that evaluates model performance while accounting for the number of interaction rounds required to reach conclusions. Moreover, SE Arena incorporates a new feature called RepoChat, which automatically injects repository-related context (e.g., issues, commits, pull requests) into the conversation, further aligning evaluations with real-world development processes. This paper outlines the design and capabilities of SE Arena, emphasizing its potential to advance the evaluation and practical application of FMs in software engineering.
Recent advancements in Large Language Models (LLMs) have paved the way for Large Code Models (LCMs), enabling automation in complex software engineering tasks, such as code generation, software testing, and program comprehension, among others. Tools like GitHub Copilot and ChatGPT have shown substantial benefits in supporting developers across various practices. However, the ambition to scale these models to trillion-parameter sizes, exemplified by GPT-4, poses significant challenges that limit the usage of Artificial Intelligence (AI)-based systems powered by large Deep Learning (DL) models. These include rising computational demands for training and deployment and issues related to trustworthiness, bias, and interpretability. Such factors can make managing these models impractical for many organizations, while their "black-box'' nature undermines key aspects, including transparency and accountability. In this paper, we question the prevailing assumption that increasing model parameters is always the optimal path forward, provided there is sufficient new data to learn additional patterns. In particular, we advocate for a Neurosymbolic research direction that combines the strengths of existing DL techniques (e.g., LLMs) with traditional symbolic methods--renowned for their reliability, speed, and determinism. To this end, we outline the core features and present preliminary results for our envisioned approach, aimed at establishing the first Neurosymbolic Program Comprehension (NsPC) framework to aid in identifying defective code components.
Existing reinforcement learning strategies based on outcome supervision have proven effective in enhancing the performance of large language models(LLMs) for code generation. While reinforcement learning based on process supervision has shown great promise in handling multi-step reasoning tasks, its effectiveness in code generation remains largely underexplored and underjustified. The primary obstacle stems from the resource-intensive nature of constructing high-quality process-supervised data, which demands substantial human expertise and computational resources. In response to this challenge, we propose a "statement mutation/refactoring-compile and execution verification" strategy: mutating and refactoring code line-by-line through a teacher model, and utilizing compiler execution results to automatically label each line, resulting in line-by-line process-supervised data, which is pivotal for training a process-supervised reward model. The trained reward model is then integrated into the PRLCoder framework, followed by experimental validation on several benchmarks. Experimental results demonstrate that process-supervised reinforcement learning significantly surpasses methods relying solely on outcome supervision. Notably, in tackling complex code generation tasks, process-supervised reinforcement learning shows a clear advantage, ensuring both the integrity of the code generation process and the correctness of the generation results.
Large Language Models (LLMs) are becoming increasingly competent across various domains, educators are showing a growing interest in integrating these LLMs into the learning process. Especially in software engineering, LLMs have demonstrated qualitatively better capabilities in code summarization, code generation, and debugging. Despite various research on LLMs for software engineering tasks in practice, limited research captures the benefits of LLMs for pedagogical advancements and their impact on the student learning process. To this extent, we analyze 126 undergraduate students' interaction with an AI assistant during a 13-week semester to understand the benefits of AI for software engineering learning. We analyze the conversations, code generated, code utilized, and the human intervention levels to integrate the code into the code base. Our findings suggest that students prefer ChatGPT over CoPilot. Our analysis also finds that ChatGPT generates responses with lower computational complexity compared to CoPilot. Furthermore, conversational-based interaction helps improve the quality of the code generated compared to auto-generated code. Early adoption of LLMs in software engineering is crucial to remain competitive in the rapidly developing landscape. Hence, the next generation of software engineers must acquire the necessary skills to interact with AI to improve productivity.
Since time immemorial an old adage has always seemed to ring true: you cannot use a high-level productive programming language like Python or R for real-time control and embedded-systems programming, you must rewrite your program in C. We present a counterexample to this mantra by demonstrating how recent compiler developments in the Julia programming language allow users of Julia and the equation-based modeling language ModelingToolkit to compile and deploy binaries for real-time model-based estimation and control. Contrary to the approach taken by a majority of modeling and simulation tools, we do not generate C code, and instead demonstrate how we may use the native Julia code-generation pipeline through LLVM to compile architecture-specific binaries from high-level code. This approach avoids many of the restrictions typically placed on high-level languages to enable C-code generation. As case studies, we include a nonlinear state estimator derived from an equation-based model which is compiled into a program that performs state estimation for deployment onto a Raspberry Pi, as well as a PID controller library implemented in Julia and compiled into a shared library callable from a C program.
Recent work has shown that Large Language Models (LLMs) are not only a suitable tool for code generation but also capable of generating annotation-based code specifications. Scaling these methodologies may allow us to deduce provable correctness guarantees for large-scale software systems. In comparison to other LLM tasks, the application field of deductive verification has the notable advantage of providing a rigorous toolset to check LLM-generated solutions. This short paper provides early results on how this rigorous toolset can be used to reliably elicit correct specification annotations from an unreliable LLM oracle.
Artificial Intelligence (AI)-driven code generation tools are increasingly used throughout the software development lifecycle to accelerate coding tasks. However, the security of AI-generated code using Large Language Models (LLMs) remains underexplored, with studies revealing various risks and weaknesses. This paper analyzes the security of code generated by LLMs across different programming languages. We introduce a dataset of 200 tasks grouped into six categories to evaluate the performance of LLMs in generating secure and maintainable code. Our research shows that while LLMs can automate code creation, their security effectiveness varies by language. Many models fail to utilize modern security features in recent compiler and toolkit updates, such as Java 17. Moreover, outdated methods are still commonly used, particularly in C++. This highlights the need for advancing LLMs to enhance security and quality while incorporating emerging best practices in programming languages.
This paper introduces a hierarchical framework for decision-making and optimization, leveraging Large Language Models (LLMs) for adaptive code generation. Instead of direct decision-making, LLMs generate and refine executable control policies through a meta-policy that guides task generation and a base policy for operational actions. Applied to a simplified microgrid scenario, the approach achieves up to 15 percent cost savings by iteratively improving battery control strategies. The proposed methodology lays a foundation for integrating LLM-based tools into planning and control tasks, offering adaptable and scalable solutions for complex systems while addressing challenges of uncertainty and reproducibility.
The majority of research around Large Language Models (LLM) application to software development has been on the subject of code generation. There is little literature on LLMs' impact on requirements engineering (RE), which deals with the process of developing and verifying the system requirements. Within RE, there is a subdiscipline of requirements elicitation, which is the practice of discovering and documenting requirements for a system from users, customers, and other stakeholders. In this analysis, we compare LLM's ability to elicit requirements of a software system, as compared to that of a human expert in a time-boxed and prompt-boxed study. We found LLM-generated requirements were evaluated as more aligned (+1.12) than human-generated requirements with a trend of being more complete (+10.2%). Conversely, we found users tended to believe that solutions they perceived as more aligned had been generated by human experts. Furthermore, while LLM-generated documents scored higher and performed at 720x the speed, their cost was, on average, only 0.06% that of a human expert. Overall, these findings indicate that LLMs will play an increasingly important role in requirements engineering by improving requirements definitions, enabling more efficient resource allocation, and reducing overall project timelines.
Large Language Models (LLMs) have become an essential tool in the programmer's toolkit, but their tendency to hallucinate code can be used by malicious actors to introduce vulnerabilities to broad swathes of the software supply chain. In this work, we analyze package hallucination behaviour in LLMs across popular programming languages examining both existing package references and fictional dependencies. By analyzing this package hallucination behaviour we find potential attacks and suggest defensive strategies to defend against these attacks. We discover that package hallucination rate is predicated not only on model choice, but also programming language, model size, and specificity of the coding task request. The Pareto optimality boundary between code generation performance and package hallucination is sparsely populated, suggesting that coding models are not being optimized for secure code. Additionally, we find an inverse correlation between package hallucination rate and the HumanEval coding benchmark, offering a heuristic for evaluating the propensity of a model to hallucinate packages. Our metrics, findings and analyses provide a base for future models, securing AI-assisted software development workflows against package supply chain attacks.
Large language models (LLMs) face significant token efficiency bottlenecks in code generation and logical reasoning tasks, a challenge that directly impacts inference cost and model interpretability. This paper proposes a formal framework based on symbolic compression,integrating combinatory logic, information-theoretic optimal encoding, and context-aware inference techniques to achieve a step-change improvement in token efficiency while preserving semantic integrity. We establish a mathematical framework within a functional programming paradigm, derive the quantitative relationship between symbolic density and model interpretability, and propose a differentiable compression factor metric to evaluate encoding efficiency. Furthermore, we leverage parameter-efficient fine-tuning (PEFT) techniques to achieve a low-cost application of the GAEL language. Experimental results show that this method achieves a 78.3% token compression rate in code generation tasks while improving logical traceability by 62% through structural explicitness. This research provides new theoretical tools for efficient inference in LLMs and opens a symbolic path for modelinterpretability research.
Large language models based Multi Agent Systems (MAS) have demonstrated promising performance for enhancing the efficiency and accuracy of code generation tasks. However,most existing methods follow a conventional sequence of planning, coding, and debugging,which contradicts the growth-driven nature of human learning process. Additionally,the frequent information interaction between multiple agents inevitably involves high computational costs. In this paper,we propose Cogito,a neurobiologically inspired multi-agent framework to enhance the problem-solving capabilities in code generation tasks with lower cost. Specifically,Cogito adopts a reverse sequence: it first undergoes debugging, then coding,and finally planning. This approach mimics human learning and development,where knowledge is acquired progressively. Accordingly,a hippocampus-like memory module with different functions is designed to work with the pipeline to provide quick retrieval in similar tasks. Through this growth-based learning model,Cogito accumulates knowledge and cognitive skills at each stage,ultimately forming a Super Role an all capable agent to perform the code generation task. Extensive experiments against representative baselines demonstrate the superior performance and efficiency of Cogito. The code is publicly available at https://anonymous.4open.science/r/Cogito-0083.
The evaluation of generative or discriminative large language model (LLM)-based systems is often a complex multi-dimensional problem. Typically, a set of system configuration alternatives are evaluated on one or more benchmark datasets, each with one or more evaluation metrics, which may differ between datasets. We often want to evaluate -- with a statistical measure of significance -- whether systems perform differently either on a given dataset according to a single metric, on aggregate across metrics on a dataset, or across datasets. Such evaluations can be done to support decision-making, such as deciding whether a particular system component change (e.g., choice of LLM or hyperparameter values) significantly improves performance over the current system configuration, or, more generally, whether a fixed set of system configurations (e.g., a leaderboard list) have significantly different performances according to metrics of interest. We present a framework implementation that automatically performs the correct statistical tests, properly aggregates the statistical results across metrics and datasets (a nontrivial task), and can visualize the results. The framework is demonstrated on the multi-lingual code generation benchmark CrossCodeEval, for several state-of-the-art LLMs.
The irruption of DeepSeek-R1 constitutes a turning point for the AI industry in general and the LLMs in particular. Its capabilities have demonstrated outstanding performance in several tasks, including creative thinking, code generation, maths and automated program repair, at apparently lower execution cost. However, LLMs must adhere to an important qualitative property, i.e., their alignment with safety and human values. A clear competitor of DeepSeek-R1 is its American counterpart, OpenAI's o3-mini model, which is expected to set high standards in terms of performance, safety and cost. In this technical report, we systematically assess the safety level of both DeepSeek-R1 (70b version) and OpenAI's o3-mini (beta version). To this end, we make use of our recently released automated safety testing tool, named ASTRAL. By leveraging this tool, we automatically and systematically generated and executed 1,260 test inputs on both models. After conducting a semi-automated assessment of the outcomes provided by both LLMs, the results indicate that DeepSeek-R1 produces significantly more unsafe responses (12%) than OpenAI's o3-mini (1.2%).