SELAB

SELAB RESEARCH

Trustworthy AI-Assisted Software Engineering

We study whether the tests, gates, and benchmarks software engineering relies on still hold when the code is written by AI.

Research Evolution

Research Trajectory

Our current Agentic AI agenda builds on earlier work in software analytics, deep learning, repository mining, and program repair.

Trustworthy AI-Assisted Software Engineering

Ongoing study of measurement validity, residual failure after passing tests, persistent unsafe artifacts from coding agents, complementarity of assurance gates, and whether empirical conclusions survive model generations.

  1. MEASUREMENT VALIDITY IN PROGRESS

    Evaluation Protocol Validity

    Whether an evaluation protocol measures agent capability at all, and whether protocol choice alone can reverse a published conclusion. Examined through repeated runs and protocol transfer.

    • Multi-Run Evaluation
    • Protocol Transfer
    • Reproduction

  2. RESIDUAL RISK

    Residual Failures Behind Green Tests

    How much behavioral failure remains in AI-generated patches that pass every test, and under what conditions an acceptance test can count as reliability evidence.

    • Behavioral Divergence
    • Lower-Bound Estimation
    • Reliability Evidence

  3. AI SOFTWARE SECURITY IN PROGRESS

    Persistent Unsafe Artifacts

    Whether repository-level prompt injection ends at attack success or leaves durable unsafe artifacts in the codebase, and whether those artifacts pass an ordinary development pipeline.

    • Repository Injection
    • Artifact Persistence
    • Pipeline Escape

  4. SOFTWARE ASSURANCE

    Complementarity of Assurance Gates

    Whether tests, static analysis, and CI checks detect different defects or the same ones, and which combinations justify their cost inside an assurance pipeline.

    • Conditional Detection
    • Gate Complementarity
    • Cost–Benefit

  5. TEMPORAL VALIDITY

    Longitudinal Reproduction

    Whether a conclusion obtained today survives a change of model generation, and how model drift bounds the useful lifetime of an empirical finding.

    • Model Drift
    • Replication
    • Temporal Validity
Agentic AI for Software Engineering

Repository-grounded agents, tool-using workflows, executable verification, reliability evaluation, and evidence-aware coordination for software maintenance.

  • Agentic Repository Intelligence
  • Tool-Using Agents
  • Reliable Software Change
  • Agent Evaluation
  • Multi-Agent Coordination
  • Human Oversight
Repository-Aware LLM Systems

Retrieval-augmented reasoning, structure-aware repository context, security-bug analysis, feature selection, and duplicate detection.

  • RAG for Code
  • AST Retrieval
  • Security Bug Analysis
  • Feature Selection
  • Duplicate Detection
  • Small LLM Workflows
Applied AI & Software Mining

Program repair, bug severity and triage, clinical NLP, voice-phishing detection, and emotion-aware applications.

  • CodeBERT APR
  • Clinical NLP
  • Bug Triage
  • Voice Phishing Detection
  • Emotion AI
Deep Learning for Software Engineering

Deep learning for bug triage, duplicate detection, bug localization, and program repair.

  • CNN-LSTM Triage
  • BERT Duplicate Detection
  • SeqGAN Repair
  • Bug Localization
Software Analytics Foundations

Bug-report mining, developer recommendation, severity prediction, topic modeling, and social-network analysis.

  • Topic Modeling
  • Social Network Analysis
  • GAN-based Repair
  • Software Analytics