Large language models (LLMs) have shown exceptional performance in code generation and understanding tasks, yet their high computational costs hinder broader adoption. One important factor is the inherent verbosity of programming languages, such as unnecessary formatting elements and lengthy boilerplate code. This leads to inflated token counts in both input and generated outputs, which increases inference costs and slows down the generation process. Prior work improves this through simplifying programming language grammar, reducing token usage across both code understanding and generation tasks. However, it is confined to syntactic transformations, leaving significant opportunities for token reduction unrealized at the semantic level.In this work, we propose Token Sugar, a concept that replaces frequent and verbose code patterns with reversible, token-efficient shorthand in the source code. To realize this concept in practice, we designed a systematic solution that mines high-frequency, token-heavy patterns from a code corpus, maps each to a unique shorthand, and integrates them into LLM pretraining via code transformation. With this solution, we obtain 799 (code pattern, shorthand) pairs, which can reduce up to 15.1% token count in the source code and is complementary to existing syntax-focused methods. We further trained three widely used LLMs on Token Sugar-augmented data. Experimental results show that these models not only achieve significant token savings (up to 11.2% reduction) during generation but also maintain near-identical Pass@1 scores compared to baselines trained on unprocessed code.

Currently, large pre-trained language models are widely applied in neural code completion systems. Though large code models significantly outperform their smaller counterparts, around 70% of displayed code completions from Github Copilot are not accepted by developers. Being reviewed but not accepted, their help to developer productivity is considerably limited and may conversely aggravate the workload of developers, as the code completions are automatically and actively generated in state-of-the-art code completion systems as developers type out once the service is enabled. Even worse, considering the high cost of the large code models, it is a huge waste of computing resources and energy, which severely goes against the sustainable development principle of AI technologies. However, such waste has never been realized, not to mention effectively addressed, in the research community for neural code completion. Hence, preventing such unhelpful code completions from happening in a cost-friendly way is of urgent need. To fill this significant gap, we first investigate the prompts of unhelpful code completions, called “low-return prompts.” We empirically identify four observable patterns in low-return prompts, each lacking necessary information, making it difficult to address through enhancements to the model’s accuracy alone. This demonstrates the feasibility of identifying such low-return prompts based on the prompts themselves. Motivated by this finding, we propose an early-rejection mechanism to turn down low-return prompts by foretelling the code completion qualities. The prompts that are estimated to receive unhelpful code completions will not be sent to the model. Furthermore, we investigated five types of estimators to demonstrate the feasibility of the mechanism. The experimental results show that the estimator can reject 20% of code completion requests with a 97.4% precision. To the best of our knowledge, it is the first systemic approach to address the problem of unhelpful code completions and this work also sheds light on an important research direction of large code models.

Leveraging recent advancements in large language models, modern neural code completion models have demonstrated the capability to generate highly accurate code suggestions. However, their massive size poses challenges in terms of computational costs and environmental impact, hindering their widespread adoption in practical scenarios. Dynamic inference emerges as a promising solution, as it allocates minimal computation during inference while maintaining the model's performance. In this research, we explore dynamic inference within the context of code completion. Initially, we conducted an empirical investigation on GPT-2, focusing on the inference capabilities of intermediate layers for code completion. We found that 54.4% of tokens can be accurately generated using just the first layer, signifying significant computational savings potential. Moreover, despite using all layers, the model still fails to predict 14.5% of tokens correctly, and the subsequent completions continued from them are rarely considered helpful, with only a 4.2% Acceptance Rate. These findings motivate our exploration of dynamic inference in code completion and inspire us to enhance it with a decision-making mechanism that stops the generation of incorrect code. We thus propose a novel dynamic inference method specifically tailored for code completion models. This method aims not only to produce correct predictions with largely reduced computation but also to prevent incorrect predictions proactively. Our extensive evaluation shows that it can averagely skip 1.7 layers out of 16 layers in the models, leading to an 11.2% speedup with only a marginal 1.1% reduction in ROUGE-L.

Neural code generation systems have recently attracted increasing attention to improve developer productivity and speed up software development. Typically, these systems maintain a pre-trained neural model and make it available to general users as a service (e.g., through remote APIs) and incorporate a feedback mechanism to extensively collect and utilize the users' reaction to the generated code, i.e., user feedback. However, the security implications of such feedback have not yet been explored. With a systematic study of current feedback mechanisms, we find that feedback makes these systems vulnerable to feedback data injection (FDI) attacks. We discuss the methodology of FDI attacks and present a pre-attack profiling strategy to infer the attack constraints of a targeted system in the black-box setting. We demonstrate two proof-of-concept examples utilizing the FDI attack surface to implement prompt injection attacks and backdoor attacks on practical neural code generation systems. The attacker may stealthily manipulate a neural code generation system to generate code with vulnerabilities, attack payload, and malicious and spam messages. Our findings reveal the security implications of feedback mechanisms in neural code generation systems, paving the way for increasing their security.

Artificial Intelligence (AI) models have emerged as another important audience for programming languages alongside humans and machines, as we enter the era of large language models (LLMs). LLMs can now perform well in coding competitions and even write programs like developers to solve various tasks, including mathematical problems. However, the grammar and layout of current programs are designed to cater the needs of human developers -- with many grammar tokens and formatting tokens being used to make the code easier for humans to read. While this is helpful, such a design adds unnecessary computational work for LLMs, as each token they either use or produce consumes computational resources. To improve inference efficiency and reduce computational costs, we propose the concept of AI-oriented grammar.This aims to represent code in a way that better suits the working mechanism of AI models. Code written with AI-oriented grammar discards formats and uses a minimum number of tokens to convey code semantics effectively. To demonstrate the feasibility of this concept, we explore and implement the first AI-oriented grammar for Python, named Simple Python (SimPy). SimPy is crafted by revising the original Python grammar through a series of heuristic rules. Programs written in SimPy maintain identical Abstract Syntax Tree (AST) structures to those in standard Python. This allows for not only execution via a modified AST parser, but also seamless transformation between programs written in Python and SimPy, enabling human developers and LLMs to use Python and SimPy, respectively, when they need to collaborate. We also look into methods to help existing LLMs understand and use SimPy effectively. In the experiments, compared with Python, SimPy enables a reduction in token usage by 13.5% and 10.4% for CodeLlama and GPT-4, respectively, when completing the same set of code-related tasks. Additionally, these models can maintain or even improve their performance when using SimPy instead of Python for these tasks. With these promising results, we call for further contributions to the development of AI-oriented program grammar within our community.

Abstract Recent studies on adversarial examples expose vulnerabilities of natural language processing models. Existing techniques for generating adversarial examples are typically driven by deterministic hierarchical rules that are agnostic to the optimal adversarial examples, a strategy that often results in adversarial samples with a suboptimal balance between magnitudes of changes and attack successes. To this end, in this research we propose two algorithms, Reversible Jump Attack (RJA) and Metropolis–Hasting Modification Reduction (MMR), to generate highly effective adversarial examples and to improve the imperceptibility of the examples, respectively. RJA utilizes a novel randomization mechanism to enlarge the search space and efficiently adapts to a number of perturbed words for adversarial examples. With these generated adversarial examples, MMR applies the Metropolis–Hasting sampler to enhance the imperceptibility of adversarial examples. Extensive experiments demonstrate that RJA-MMR outperforms current state-of-the-art methods in attack performance, imperceptibility, fluency and grammar correctness.

Recent research has revealed that natural language processing (NLP) models are vulnerable to adversarial examples. However, the current techniques for generating such examples rely on deterministic heuristic rules, which fail to produce optimal adversarial examples. In response, this study proposes a new method called the Fraud's Bargain Attack (FBA), which uses a randomization mechanism to expand the search space and produce high-quality adversarial examples with a higher probability of success. FBA uses the Metropolis-Hasting sampler, a type of Markov Chain Monte Carlo sampler, to improve the selection of adversarial examples from all candidates generated by a customized stochastic process called the Word Manipulation Process (WMP). The WMP method modifies individual words in a contextually-aware manner through insertion, removal, or substitution. Through extensive experiments, this study demonstrates that FBA outperforms other methods in terms of attack success rate, imperceptibility and sentence quality.

Code search, which consists in retrieving relevant code snippets from a codebase based on a given query, provides developers with useful references during software development. Over the years, techniques alternatively adopting different mechanisms to compute the relevance score between a query and a code snippet have been proposed to advance the state of the art in this domain, including those relying on information retrieval, supervised learning, and pre-training. Despite that, the usefulness of existing techniques is still compromised since they cannot effectively handle all the diversified queries and code in practice. To tackle this challenge, we presentDancer, a data fusion based code searcher. Our intuition (also the basic hypothesis of this study) is that existing techniques may complement each other because of the intrinsic differences in their working mechanisms. We have validated this hypothesis via an exploratory study. Based on that, we propose to fuse the results generated by different code search techniques so that the advantage of each standalone technique can be fully leveraged. Specifically, we treat each technique as a retrieval system and leverage well-known data fusion approaches to aggregate the results from different systems. We evaluate six existing code search techniques on two large-scale datasets, and exploit eight classic data fusion approaches to incorporate their results. Our experiments show that the best fusion approach is able to outperform the standalone techniques by 35% - 550% and 65% - 825% in terms of MRR (mean reciprocal rank) on the two datasets, respectively.

Recent studies on adversarial examples expose vulnerabilities of natural language processing (NLP) models. Existing techniques for generating adversarial examples are typically driven by deterministic heuristic rules that are agnostic to the optimal adversarial examples, a strategy that often results in attack failures. To this end, this research proposes Fraud's Bargain Attack (FBA), which utilizes a novel randomization mechanism to enlarge the searching space and enables high-quality adversarial examples to be generated with high probabilities. FBA applies the Metropolis-Hasting algorithm to enhance the selection of adversarial examples from all candidates proposed by a customized Word Manipulation Process (WMP). WMP perturbs one word at a time via insertion, removal, or substitution in a contextual-aware manner. Extensive experiments demonstrate that FBA outperforms the baselines in terms of attack success rate and imperceptibility.

Code datasets are of immense value for training neural-network-based code completion models, where companies or organizations have made substantial investments to establish and process these datasets. Unluckily, these datasets, either built for proprietary or public usage, face the high risk of unauthorized exploits, resulting from data leakages, license violations, etc. Even worse, the "black-box" nature of neural models sets a high barrier for externals to audit their training datasets, which further connives these unauthorized usages. Currently, watermarking methods have been proposed to prohibit inappropriate usage of image and natural language datasets. However, due to domain specificity, they are not directly applicable to code datasets, leaving the copyright protection of this emerging and important field of code data still exposed to threats. To fill this gap, we propose a method, named CodeMark, to embed user-defined imperceptible watermarks into code datasets to trace their usage in training neural code completion models. CodeMark is based on adaptive semantic-preserving transformations, which preserve the exact functionality of the code data and keep the changes covert against rule-breakers. We implement CodeMark in a toolkit and conduct an extensive evaluation of code completion models. CodeMark is validated to fulfill all desired properties of practical watermarks, including harmlessness to model accuracy, verifiability, robustness, and imperceptibility.

A longstanding dream in software engineering research is to devise effective approaches for automating development tasks based on developers' informally-specified intentions. Such intentions are generally in the form of natural language descriptions. In recent literature, a number of approaches have been proposed to automate tasks such as code search and even code generation based on natural language inputs. While these approaches vary in terms of technical designs, their objective is the same: transforming a developer's intention into source code. The literature, however, lacks a comprehensive understanding towards the effectiveness of existing techniques as well as their complementarity to each other. We propose to fill this gap through a large-scale empirical study where we systematically evaluate natural language to code techniques. Specifically, we consider six state-of-the-art techniques targeting code search, and four targeting code generation. Through extensive evaluations on a dataset of 22K+ natural language queries, our study reveals the following major findings: (1) code search techniques based on model pre-training are so far the most effective while code generation techniques can also provide promising results; (2) complementarity widely exists among the existing techniques; and (3) combining the ten techniques together can enhance the performance for 35% compared with the most effective standalone technique. Finally, we propose a post-processing strategy to automatically integrate different techniques based on their generated code. Experimental results show that our devised strategy is both effective and extensible.

Automatically transforming developers' natural language descriptions into source code has been a longstanding goal in software engineering research. Two types of approaches have been proposed in the literature to achieve this: code generation, which involves generating a new code snippet, and code search, which involves reusing existing code. However, despite existing efforts, the effectiveness of the state-of-the-art techniques remains limited. To seek for further advancement, our insight is that code generation and code search can help overcome the limitation of each other: the code generator can benefit from feedback on the quality of its generated code, which can be provided by the code searcher, while the code searcher can benefit from the additional training data augmented by the code generator to better understand code semantics. Drawing on this insight, we propose a novel approach that combines code generation and code search techniques using a generative adversarial network (GAN), enabling mutual improvement through the adversarial training. Specifically, we treat code generation and code search as the generator and discriminator in the GAN framework, respectively, and incorporate several customized designs for our tasks. We evaluate our approach in eight different settings, and consistently observe significant performance improvements for both code generation and code search. For instance, when using NatGen, a state-of-the-art code generator, as the generator and GraphCodeBERT, a state-of-the-art code searcher, as the discriminator, we achieve a 32% increase in CodeBLEU score for code generation, and a 12% increase in mean reciprocal rank for code search on a large-scale Python dataset, compared to their original performances.

The performance of neural code search is significantly influenced by the quality of the training data from which the neural models are derived. A large corpus of high-quality query and code pairs is demanded to establish a precise mapping from the natural language to the programming language. Due to the limited availability, most widely-used code search datasets are established with compromise, such as using code comments as a replacement of queries. Our empirical study on a famous code search dataset reveals that over one-third of its queries contain noises that make them deviate from natural user queries. Models trained through noisy data are faced with severe performance degradation when applied in real-world scenarios. To improve the dataset quality and make the queries of its samples semantically identical to real user queries is critical for the practical usability of neural code search. In this paper, we propose a data cleaning framework consisting of two subsequent filters: a rule-based syntactic filter and a model-based semantic filter. This is the first framework that applies semantic query cleaning to code search datasets. Experimentally, we evaluated the effectiveness of our framework on two widely-used code search models and three manually-annotated code retrieval benchmarks. Training the popular DeepCS model with the filtered dataset from our framework improves its performance by 19.2% MRR and 21.3% [email protected], on average with the three validation benchmarks.

Github Copilot, trained on billions of lines of public code, has recently become the buzzword in the computer science research and practice community. Although it is designed to help developers implement safe and effective code with powerful intelligence, practitioners and researchers raise concerns about its ethical and security problems, e.g., should the copyleft licensed code be freely leveraged or insecure code be considered for training in the first place? These problems pose a significant impact on Copilot and other similar products that aim to learn knowledge from large-scale open-source code through deep learning models, which are inevitably on the rise with the fast development of artificial intelligence. To mitigate such impacts, we argue that there is a need to invent effective mechanisms for protecting open-source code from being exploited by deep learning models. Here, we design and implement a prototype, CoProtector, which utilizes data poisoning techniques to arm source code repositories for defending against such exploits. Our large-scale experiments empirically show that CoProtector is effective in achieving its purpose, significantly reducing the performance of Copilot-like deep learning models while being able to stably reveal the secretly embedded watermark backdoors.

Third-party libraries are crucial to the development of software projects. To get suitable libraries, developers need to search through millions of libraries by filtering, evaluating, and comparing. The vast number of libraries places a barrier for programmers to locate appropriate ones. To help developers, researchers have proposed automated approaches to recommend libraries based on library usage pattern. However, these prior studies can not sufficiently match user requirements and suffer from cold-start problem. In this work, we would like to make recommendations based on requirement descriptions to avoid these problems. To this end, we propose a novel neural approach called Req2Lib which recommends libraries given descriptions of the project requirement. We use a Sequence-to-Sequence model to learn the library linked-usage information and semantic information of requirement descriptions in natural language. Besides, we apply a domain-specific pre-trained word2vec model for word embedding, which is trained over textual corpus from Stack Overflow posts. In the experiment, we train and evaluate the model with data from 5,625 java projects. Our preliminary evaluation demonstrates that Req2Lib can recommend libraries accurately.