TraceSynth uses diffusion models to generate synthetic kernel traces for ML diagnostics

Researchers have developed TraceSynth, a novel diffusion-based framework designed to generate synthetic kernel execution traces. These synthetic traces aim to supplement limited real-world data for machine learning tasks in system diagnostics. TraceSynth utilizes a Transformer-based denoising diffusion process, incorporating constraint-guided repair to maintain system invariants and models traces as multi-channel sequences. The framework demonstrates strong performance, particularly for deterministic, compute-heavy workloads like scimark2, where synthetic augmentation achieved an 87.2% F1-Macro score. The research indicates that context length is a critical factor for trace quality, and lightweight models can retain significant performance at reduced computational cost. AI

IMPACT Enables more robust and cost-effective machine learning for system diagnostics by augmenting limited real-world trace data.

RANK_REASON The cluster contains an academic paper detailing a new method for generating synthetic data using diffusion models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• ScienceCast
• scimark2
• TraceSynth
• Transformer++

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →