New Benchmark Dataset Enables Detection of AI-Generated Code in Mixed Human-AI Codebases
Researchers introduced HybridCodeAuthorship, a benchmark dataset designed to detect AI-generated code at the line level within codebases that mix human and AI-authored code. The dataset addresses a gap in existing benchmarks by using realistic industry code patterns rather than academic problems, reflecting how AI code assistants are actually used in practice. This capability is important for risk management and productivity analysis as AI-generated code becomes increasingly prevalent in software development.
A new research paper on arXiv presents HybridCodeAuthorship, a benchmark dataset for detecting AI-generated code within mixed human-AI codebases at fine-grained levels. The researchers constructed the dataset using CodeSearchNet, drawing from open-source repositories on GitHub, to create Python code files with interleaved human- and AI-authored lines that reflect authentic usage patterns of AI code assistants. The study benchmarked two state-of-the-art detection algorithms, finding that the best performer (AIGCode Detector) achieved an F1 score of 0.56 on line-level detection and 0.48 on chunk-level detection, indicating the task remains challenging. The dataset addresses a critical gap in existing benchmarks, which typically assume code is either entirely human- or entirely AI-authored and rely on academic or LeetCode-style problems rather than realistic industry code. The researchers argue this capability is essential for organizations to manage risks and analyze productivity as AI code assistants become more widely adopted.
What's missing
The study does not discuss potential limitations of the detection approach, such as how performance might vary across different LLM architectures, programming languages beyond Python, or obfuscation techniques that could evade detection. Additionally, the paper does not address the ethical or legal implications of detecting AI-generated code in proprietary codebases.
What different sources said
- arXiv cs.AICenter
HybridCodeAuthorship: A Benchmark Dataset for Line-Level Code Authorship Detection
Related
Topology-Aware Thermodynamics Improves DNA Probe Specificity Design
Researchers developed a new framework for designing DNA probes that accounts for the spatial organization of matched sequences, not just overall thermodynamic stability. Traditional methods rely on scalar measures like melting temperature and free energy, which miss how mismatches are distributed along the probe. The approach could improve diagnostic accuracy in applications like HPV detection and gene expression profiling.
Study Identifies Optimal Thermal Dose for Combining Focused Ultrasound with Immunotherapy in Tumors
Researchers used multimodal PET imaging to identify an optimal thermal dose range for focused ultrasound ablation that destroys tumor tissue while preserving conditions for immunotherapy delivery. The study found that excessive heating collapses blood vessels needed for antibody access, while insufficient heating fails to adequately reduce tumor burden. The findings could guide clinical design of combination treatments pairing thermal ablation with immunotherapies.
Plant MSH1 Protein Functions as Mismatch-Directed Nuclease for Organelle Genome Maintenance
Researchers have identified the precise mechanism by which the AtMSH1 protein in Arabidopsis plants recognizes and cleaves DNA mismatches and lesions, preventing mutations in organellar genomes. The protein combines a DNA mismatch recognition module with a nuclease domain that makes staggered cuts at specific positions relative to DNA damage. This discovery explains how plants maintain unusually low mutation rates in their mitochondrial and chloroplast DNA compared to other eukaryotes.