MindForge pipeline trains small LLMs for full software engineering lifecycle

Researchers have developed MindForge, an automated pipeline designed to train smaller language models in comprehensive software engineering tasks. This system converts open-source command-line programs into source-free training environments, providing only compiled executables and documentation. By fine-tuning the Qwen3.6-27B model on synthesized program trajectories generated by GLM-5.2, the researchers significantly improved its performance on the ProgramBench benchmark, achieving results comparable to larger frontier models. The fine-tuned model also demonstrated broad improvements across various unseen software engineering tasks, including code generation, bug fixing, and cross-language issue resolution. AI

IMPACT Enhances the capabilities of smaller language models in complex software engineering tasks, potentially lowering the barrier to entry for AI-assisted development.

RANK_REASON This is a research paper detailing a new method and model fine-tuning for software engineering tasks.

Read on Hugging Face Daily Papers →

• Deepsweg
• FeatBench
• GLM-5.2
• NL2Repo-Bench
• ProgramBench
• Qwen3.6-27B
• RepoZero-C2Rust
• SWE-bench Multilingual
• SWE Bench Pro
• SWE-bench Verified
• Claude Opus 4.7
• DeepSeek V4 Pro
• GLM-5.1

• model release
• paper
• product

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →