Search papers, labs, and topics across Lattice.
We introduce EditHero, the first benchmark for long-horizon, part-level 3D editing with natural-language instructions and target images, addressing the gap in evaluating iterative asset creation. A deterministic assembly engine ensures precise target states after each edit, with all sequences manually verified. Comparing non-agentic and LLM/VLM agentic approaches, we find agents follow instructions more closely and preserve unedited parts better, though at a significant computational cost.
LLM/VLM agents significantly outperform non-agentic methods in preserving unedited parts during iterative 3D asset modification, but current agentic edits are prohibitively slow.
3D editing methods are usually tested on a single edit, yet an asset is built through a long sequence of revisions, each of which must implement the requested change while leaving everything else unchanged. We introduce EditHero, to our knowledge the first benchmark for long-horizon, part-level 3D editing, with natural-language instructions and target images for both geometry and texture. A deterministic assembly engine produces the exact target after every edit, and every sequence is reviewed by hand. We use EditHero to compare 2 opposite approaches to 3D editing. Non-agentic methods operate top down, regenerating the object from a learned 3D representation and inferring what to keep. In contrast, LLM/VLM agents operate bottom up, editing through code that inspects the mesh and rewrites only the parts required by instructions. The non-agentic methods often miss the requested change and disturb regions that should stay fixed. Most LLMs follow instructions more closely, and all of them preserve the unedited parts better, but each of their edits takes minutes. We will release the engine and the edit sequences to support research on reliable iterative 3D editing.
