HF engineer trains Qwen3.5-9B with GRPO and GEPA to code Blender 3D rooms, GLM-5.3-Flash as judge
mervenoyann · x · 2026-09-07
Hugging Face engineer Merve Noyann open-sourced a research project built on TRL and OpenEnv that trains Qwen3.5-9B to write bpy code via tool calls, generating complete low-poly isometric room dioramas from blank Blender geometry, with GLM-5.3-Flash serving as judge.
Key details:
- 256 deterministic training prompts across four room families (studio, bedroom, kitchen, living room) plus a 64-prompt held-out eval set
- The policy gets four tools: executepython, inspectscene, renderscene (text metadata only) and finish
- Two-phase plan: run GEPA to iterate on rewards first (early runs produced no renders; current ones render but lack semantic quality), then 200 GRPO optimizer steps
- The first HF Job is evaluation-only and does not modify the base model
Code, scripts, Dockerfile and docs are public in the merve/blender-grpo-gepa HF bucket, with live rollouts viewable online.
Related event: HF engineers train a 9B model to build 3D rooms via Blender code(3 posts)→
More from coding & agent
- Meta FAIR's Self-Direction turns agent failures into reusable LoRA skills, +13.8pp on Terminal-Bench — karen_ullrich · 2026-10-27
- Dev ships 1000+ commits with agents by pushing straight to main, no complex orchestration — Dimillian · 2026-10-27
- AI code editor Superset rebrands as Paradev, raises $11.5M seed led by USV — massimosgrelli · 2026-10-27
- Coding Agent Deleted a Production Database During Code Freeze: Anaconda's Governance Fix — anacondainc · 2026-10-27
- Claude Projects now lets you drag threads into the message box or @ mention them for precise context — dfeinition · 2026-10-27
- Jane Street's Yinsky shares tips for making codebases agent-friendly: types, property tests, formal methods — sh_reya · 2026-10-27