Xenon Mixed Agentic Dataset: Coding Traces for Mixture-of-Experts Model Fine-Tuning

Xenon Mixed Agentic Datasetv1 is a curated mixture of coding agent traces designed for Supervised Fine-Tuning of Mixture-of-Experts architecture models. The dataset employs Strategic Language Weighting to route the best coding agent data to specific language paradigms. It was authored by el4 and last updated on July 18, 2026.

Use Cases

• Supervised Fine-Tuning of MoE models based on curated coding agent traces.
• Training model routers to specialize in language paradigms based on strategic language weighting.
• Research into multi-agent coding systems and their trace data.
• Benchmarking code generation models on specialized, routed data mixtures.

Strengths

• Dataset is specifically curated for fine-tuning Mixture-of-Experts models like Gemma-4-26B MoE.
• Employs Strategic Language Weighting, a described methodology for routing data.
• Last update timestamp is precisely recorded as 2026-07-18 04:56:01.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and license information are unknown, which may limit suitability assessment.

Provenance

Source

huggingface, author el4

Collection Method

Likely a curated mixture of traces from multiple coding agents.

Freshness

Last updated 2026-07-18 04:56:01; freshness should be verified.