AI
EditHero: A Benchmark for Long-Horizon Part-Level 3D Editing and Vibe Modeling
Publication: Hugging Face Daily Papers
Author: Not available in collected metadata
Source published: 30/09/2026, 20:00:00 UTC
Briefing created: 05/10/2026, 03:20:44 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 05/10/2026, 03:18:25 UTC.
Briefings
EditHero is the first benchmark for long-horizon, part-level 3D editing using natural language instructions. It compares two approaches: non-agentic methods that regenerate objects from learned representations versus LLM/VLM agents that edit via code. The deterministic engine produces exact targets after each edit, and all sequences are reviewed by hand. Results show non-agentic methods often fail to implement changes, while LLM agents preserve unedited parts better but take minutes per edit. The engine and edit sequences are released to support research on reliable iterative 3D editing.
Key points
- EditHero is the first benchmark for long-horizon, part-level 3D editing using natural language instructions for both geometry and texture.
- Non-agentic methods often fail to implement requested changes, whereas LLM/VLM agents preserve unedited parts better despite taking minutes per edit.
Context and limitations
The benchmark covers only part-level edits across a single long sequence; no single edit is tested, and all sequences are reviewed by hand rather than automatically validated.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 05/10/2026, 03:20:44 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeicusbrfasxwstpcgpxnnreatfce6ucymoesi6atswevoff6ljidzq