MindForge: Significantly Boosting Small Models' From-Scratch Coding Abilities
centre-for-swe · hf · 2026-07-31
While coding agents have made progress in modifying existing codebases, building complete programs from scratch remains a major challenge, with frontier models fully resolving less than 1% of ProgramBench tasks.
To address this, researchers introduced MindForge, a training pipeline that converts open-source command-line programs into source-free environments exposing only compiled executables and documentation.
Experiments show that fine-tuning a Qwen3.6-27B model on synthesis trajectories generated by a GLM-5.2 teacher agent increases its ProgramBench pass rate from 37.98% to 49.51%, rivaling larger frontier models. Furthermore, the fine-tuned model consistently improves across seven unseen software engineering benchmarks, including cross-language issue resolution, bug fixing, and repository generation.
More from coding & agent
- Code Review Trick: Turning Negative LLM Instructions into Positive Tasks — mattpocockuk · 2026-07-31
- 126-Byte Python Snippet Crashes Five Major Type Checkers — charliermarsh · 2026-07-31
- HeyGen Wins OSSCAR Q2 Emerging Award as Open Source Repo Grows 300x — toolstelegraph · 2026-07-31
- Microsoft Launches Echoverse to Train Computer-Use Agents in Realistic Environments — dejavucoder · 2026-07-31
- Prompt Engineering Trend: One-Shot is the New Zero-Shot — abacaj · 2026-07-31
- Heavy Developer Burns 2.2 Billion Tokens on Codex in a Single Month — ShanRizvi · 2026-07-31