Hugging Face Daily PapersXirui Li, Peng Shi, Mingwen DongSep 28, 20262 min readpaperadvanced

LEGO-Anything: Coding Agents for 3D Scene Reconstruction

Summary

LEGO-Anything is an Image-to-Code framework where a coding agent iteratively writes and executes Blender code to reconstruct 3D scenes from single images. It introduces LEGO-Bench for evaluation, showing GPT-6-astra performs best but struggles with scene initialization and self-evaluation, which LEGO-Plugin helps mitigate.

LEGO-Anything reconstructs 3D scenes as editable Blender code, not fixed 3D models.
LEGO-Bench is a new simulator-grounded benchmark for evaluating 3D scene recovery.
GPT-6-astra achieved 53.4% indoor and 39.6% outdoor scores on LEGO-Bench.
Key agent issues include weak initialization, regressive edits, and unreliable self-evaluation.

This work is significant for researchers and engineers developing AI agents for complex creative tasks and 3D scene understanding, offering a new paradigm for editable 3D reconstruction.

8/10

llm computer vision 3d reconstruction agents benchmarking blender

Related reading

Paint-Anything: Unified Any-Color Control for Image Generation and Editing arxiv.org
Paint-Anything introduces a unified hex‑prompt interface that lets diffusion models generate or edit images with any 24‑bit color per object. Using a mixed supervision pipeline and a new benchmark, it boosts object‑level color fidelity on FLUX.2‑4B by up to 85 % for text‑to‑image and 28 % for editing.
Hugging Face Daily Papersarxiv.orgJi Xie, Dewei Zhou, Xinyu Huang19d1 minpaper
SNAP3D: Physically Grounded 3D Parts for Assembly from a Single Image arxiv.org
The paper presents SNAP3D, a physics‑guided pipeline that converts a single image into a set of 3D parts that can be assembled without interpenetration. By using simulation‑driven connector placement and a new physics‑based evaluation, the method yields assemblies that are both geometrically accurate and stable enough for 3D printing.
Hugging Face Daily Papersarxiv.orgYu-Rou Tuan, Hao-Tang Tsui, Nicolas Ugrinovic25d1 minpaper
Octrees as an Explicit 3D Language arxiv.org
OctLLM treats 3D geometry as a sequence of octree occupancy tokens, using a Sparse Octree to keep sequences short while preserving shape. It adds lightweight 3D branches to a frozen vision‑language backbone, achieving state‑of‑the‑art image‑to‑3D generation with far fewer trainable parameters.
Hugging Face Daily Papersarxiv.orgRan Dan, Si-Tong Wei, Pengfei Xiong5d1 minpaper
Towards Self-Driving Codebases detail.dev
The post argues that AI agents could eventually handle low‑level engineering tasks—bug fixing, debugging, UI consistency, growth experiments—if the dev toolchain is made “agent‑legible”. It outlines missing primitives (global memory, code‑base rot prevention, better dev environments) and proposes a bootstrapping process to measure and improve a repo’s “agent readiness”. The piece is largely specu…
Hacker News front pagedetail.dev18d9 minHN12099
CodeMidas: Scaling Agentic Coding RL Environments from Code Itself arxiv.org
CodeMidas builds RL environments directly from open‑source code: agents explore a repo, infer a spec, generate tests from the original implementation, and filter tasks via execution checks. The pipeline yields 5,545 high‑quality coding tasks across 23 languages and 15 domains. Training the MiMo‑V2.5 agent with GRPO on this dataset improves benchmark scores by 8‑18% (e.g., DeepSWE +11.7%, ProgramB…
Hugging Face Daily Papersarxiv.orgBowen Ye, Lei Li, Shicheng Li18d1 minpaper
AutoRef: Harness Optimization for Agentic Multi-Reference Image Generation arxiv.org
AutoRef uses a coding LLM to automatically rewrite the harness that orchestrates multi‑reference image generation, keeping the generator frozen. The resulting harness lifts FLUX.2‑4B scores from 5.72 to 7.37 on MultiBanana and generalizes across models and benchmarks.
Hugging Face Daily Papersarxiv.orgYuta Oshima, Ku Onoda, Yusuke Iwasawa8d2 minpaper

LEGO-Anything: Coding Agents for 3D Scene Reconstruction

Summary

Related reading

Paint-Anything: Unified Any-Color Control for Image Generation and Editingarxiv.org

SNAP3D: Physically Grounded 3D Parts for Assembly from a Single Imagearxiv.org

Octrees as an Explicit 3D Languagearxiv.org

Towards Self-Driving Codebasesdetail.dev

CodeMidas: Scaling Agentic Coding RL Environments from Code Itselfarxiv.org

AutoRef: Harness Optimization for Agentic Multi-Reference Image Generationarxiv.org

Paint-Anything: Unified Any-Color Control for Image Generation and Editing arxiv.org

SNAP3D: Physically Grounded 3D Parts for Assembly from a Single Image arxiv.org

Octrees as an Explicit 3D Language arxiv.org

Towards Self-Driving Codebases detail.dev

CodeMidas: Scaling Agentic Coding RL Environments from Code Itself arxiv.org

AutoRef: Harness Optimization for Agentic Multi-Reference Image Generation arxiv.org