RESEARCH INITIATIVE

Methz M.1 Code

Exploring efficient adaptation of suitably licensed open-weight models for code understanding and generation.

Research focus

LoRA, QLoRA, code generation, debugging, code explanation, and efficient inference.

Proposed methodology

Establish data licensing, baseline benchmarks, reproducible fine-tuning, and safety evaluation before any release.

Roadmap

Model selection → baseline evaluation → adapter training → benchmark reporting.