Research
On-device research index

arXiv research

A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.

169,341 papers · 148 categories

Trend · papers per month

25.0%50.0%75.0%100.0% · Sep 199219922001200920182026
48 results for source code generation

Contrastive Code Representation Learning improves code summarization and type inference.

problem Code representations are sensitive to edits, hindering downstream semantic understanding tasks.
method ContraCode: a contrastive pre-training task that learns code functionality.
result Contrastive pre-training improves code summarization and type inference accuracy.

Deep neural network predicts semantic labels for source code.

problem Difficulty in labeling and understanding new programming languages and functionalities.
method Language-agnostic deep convolutional neural network trained on Stack Overflow code snippets.
result Mean area under ROC of 0.957 and top-1 accuracy of 86.6% on Github code documents.

We study the problem of building generative models of natural source code (NSC); that is, source code written and understood by humans. Our primary contribution is to describe a family of generative models for NSC that have three key properties: First, they incorporate both sequential and hierarchical structure. Second…

2014-01-02abs ↗pdf ↗

Paper presents a code authorship attribution attack using adversarial learning.

problem Misleading attribution of source code using machine learning methods.
method Exploits adversarial examples and semantics-preserving code transformations guided by Monte-Carlo tree search.
result Demonstrates substantial effect on attribution methods, reducing accuracy from over 88% to 1%.

Paper explores understanding of neural source code embeddings.

problem Lack of understanding of contents and characteristics of code2vec embeddings.
method Small case study using code2vec embeddings to create binary SVM classifiers and compare performance with handcrafted features.
result Code2vec embeddings perform similarly to handcrafted features and have more evenly distributed information gains.

Generative model uses graphs to create natural-sounding code.

problem Creating semantically meaningful source code with syntactic and semantic constraints.
method Graph representation for intermediate state, interleaves grammar-driven expansion with graph augmentation and neural message passing.
result Generative model outperforms baselines in generating natural-sounding code.

Backdoors can be implanted in neural models of source code, and we detect and remove them.

problem Vulnerability of neural models to backdoors in source code.
method Defined and implemented various backdoor classes, adapted robust statistics algorithms, and detected poisoned data through spectral signatures.
result Demonstrated the ease of injecting and removing backdoors in neural models of source code.

Convolutional neural network summarizes code comments across multiple languages.

problem Insufficient or missing comments in source code.
method Language-agnostic encoder-decoder model with open vocabulary.
result Comparable results to state-of-the-art on single-language data; first results on multi-language data.

Improves code2vec for Java classes by obfuscating variable names.

problem Code2vec's reliance on variable names makes it vulnerable to typos and attacks.
method Obfuscate variable names during code2vec training and aggregate method embeddings for class-level predictions.
result Obfuscated variable names improve model's robustness and accuracy.

TreeCaps improves code comprehension for software developers.

problem Processing code efficiently for software developers.
method Tree-based capsule networks for capturing code syntactical structures and dependencies.
result TreeCaps outperforms other approaches in classifying program functionalities.

This work proposes a meta-learning approach for better adaptation of source code models.

problem Adapting source code models to unseen local contexts.
method Formulated as a meta-learning problem, selecting targeted information for adaptation.
result Improved performance in code auto-completion tasks, especially for identifiers and literals.

The paper optimizes querying schemes for crowdsourced classification using XOR queries.

problem Optimizing querying schemes for crowdsourced classification.
method Modeling crowdsourced labeling/classification as source coding problem, leveraging connections to channel coding.
result Provides querying schemes with almost optimal number of queries, each involving a constant number of labels.

Study on how deterministic dependencies affect information synergy and redundancy.

problem Understanding how deterministic dependencies impact information synergy and redundancy.
method Systematic analysis of deterministic dependencies on information decomposition.
result Identifies how negative terms can originate from deterministic dependencies and discusses implications for neural coding.

Devign uses graph neural networks to identify vulnerabilities efficiently.

problem Challenging and tedious process of identifying vulnerabilities in software systems.
method Devign employs a graph neural network to classify graph-level vulnerabilities using comprehensive code semantic representations.
result Devign significantly outperforms state-of-the-art models in vulnerability identification.

Neural network identifies undeclared variables and infers their types.

problem Undeclared variable errors in programs.
method Trained on structural semantic details of AST, identifies and infers types of undeclared variables.
result Correctly identified and inferred types for 80% of programs with undeclared variable errors.

PATOIS synthesizes code from natural language using learned code idioms.

problem Synthesizing general-purpose source code from natural language specifications is challenging.
method PATOIS uses a neural synthesizer that interleaves high-level and low-level reasoning, incorporating learned code idioms from a corpus.
result Using learned code idioms improves the synthesizer's accuracy on semantic parsing datasets.

Model learns code representations from comments for data analysis tasks.

problem Lack of descriptive labels for analyzing large code corpora.
method Weakly supervised transformer architecture for joint code and comment representation.
result Model achieves 38% accuracy increase over expert-supplied heuristics.

Generative Adversarial Network repairs software bugs without labeled data.

problem Automated repair of software vulnerabilities.
method Adversarial learning approach mapping between source and target domains.
result Effective at repairing software vulnerabilities, close to seq2seq approaches.

This paper studies communication efficiency in federated learning by optimizing the sum-rate-distortion function for indirect multiterminal source coding.

problem Indirect multiterminal source coding in federated learning where edge devices send noisy gradients to the server.
method Analyzes the rate region for the quadratic vector Gaussian CEO problem under unbiased estimator and derives an explicit formula for the sum-rate-distortion function.
result Derives an explicit formula for the sum-rate-distortion function in the special case of identical gradients over edge devices.

Tangent automates derivatives in Python, improving expressiveness and performance.

problem Efficiently calculating derivatives for complex models in Python.
method Source-code transformation for dynamically typed array programming.
result Demonstrates improved expressiveness and performance in automatic differentiation.

SLM models code syntax as trees to generate any programming language code.

problem Generating any piece of code in a given language without restrictions.
method Structural language modeling (SLM) decomposes code into ASTs and estimates probabilities over nodes.
result SLM model generates arbitrary code in any language, outperforming previous methods.