PulseAugur
LIVE 13:45:08
research · [2 sources] ·
0
research

LLMs advance code generation with new Python corpus and circuit design framework

Researchers have introduced OpenClassGen, a substantial dataset comprising over 324,000 Python classes sourced from open-source projects, designed to facilitate LLM research in code generation. This corpus includes detailed static code metrics and self-contained class skeletons, enabling more robust evaluation than previous benchmarks. Initial tests on models like GPT-4-mini and Claude-4-Sonnet revealed strong semantic understanding but moderate functional correctness, highlighting the dataset's utility in differentiating LLM capabilities. Separately, a new framework for analog circuit sizing leverages LLM-derived equations to create interpretable and self-calibrating design functions. This approach uses a single simulation to extract process parameters and a feedback mechanism to correct analytical inaccuracies, enabling rapid convergence and cross-process portability without retraining. AI

Summary written by gemini-2.5-flash-lite from 2 sources. How we write summaries →

IMPACT New datasets and frameworks are released to improve LLM performance in code generation and circuit design.

RANK_REASON The cluster contains two academic papers detailing new datasets and frameworks for AI research.

Read on arXiv cs.AI →

COVERAGE [2]

  1. arXiv cs.AI TIER_1 · Musfiqur Rahman, SayedHassan Khatoonabadi, Emad Shihab ·

    OpenClassGen: A Large-Scale Corpus of Real-World Python Classes for LLM Research

    arXiv:2504.15564v3 Announce Type: replace-cross Abstract: Existing class-level code generation datasets are either synthetic (ClassEval: 100 classes) or insufficient in scale for modern training needs (RealClassEval: 400 classes), hindering robust evaluation and empirical analysi…

  2. arXiv cs.AI TIER_1 · Antonio J. Bujana, Aydin I. Karsilayan ·

    A Self-Calibrating Framework for Analog Circuit Sizing Using LLM-Derived Analytical Equations

    arXiv:2604.07387v2 Announce Type: replace-cross Abstract: We present a design automation framework for analog circuit sizing that produces calibrated, topology-specific analytical equations from raw circuit netlists. A large language model (LLM) derives a complete Python sizing f…