CodeAlchemy: New Framework Generates Massive Synthetic Code Training Data to Improve AI Model Performance
Researchers introduced CodeAlchemy, a synthetic data generation framework that creates 500+ billion tokens of training data from publicly sourced code across 15 languages using five transformation strategies. The framework includes novel benchmarks (DevEval and TraceEval) that reveal significant gaps in how well even frontier AI models understand code execution and semantics. The work demonstrates that smaller 3B-parameter models trained on this synthetic data can outperform much larger models, suggesting synthetic code data is a promising direction for improving code understanding in AI systems.
CodeAlchemy addresses a gap in AI code model training by generating semantically-rich synthetic data at unprecedented scale. The framework applies five strategies to transform raw code: CodeEnhance (quality improvements), CodeQA (template-based problems), CodeDev (developer tasks), CodeDialogue (multi-turn conversations), and CodeTrace (execution traces with instrumentation). Processing 3 corpora across 15 languages, the system generated over 500 billion tokens of synthetic data plus 350 billion reasoning tokens—orders of magnitude more than prior work. Notably, CodeTrace executed 1.3 million files across 14 languages and 5,000 libraries to capture control flow and library knowledge. The researchers introduced two new benchmarks: DevEval for developer tasks and TraceEval for execution prediction, finding that frontier models like Claude Sonnet 4.5 achieve only 5.6% exact match on TraceEval. Their 3B-parameter models achieved strong results (83.5% on HumanEval, 63.2% on MBPP) and outperformed models 10-32 times larger.
What's missing
The paper does not discuss potential limitations of the synthetic data approach, such as whether synthetic data might introduce biases or artifacts that differ from real-world code patterns, or how the quality and diversity of the 500B+ tokens compare to human-written code distributions. Additionally, the computational cost and environmental impact of generating and training on this scale of synthetic data is not addressed.
What different sources said
- arXiv cs.LGCenter
CodeAlchemy: Synthetic Code Rewriting at Scale
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.