Implementation Details of Tree-Diffusion: Architecture and Training for Inverse Graphics

27 Sept 2025

This article provides the technical implementation details of the Tree-Diffusion architecture using PyTorch and NF-ResNet.

The Role of Mutation Path Algorithms in Tree-Diffusion Program Synthesis

27 Sept 2025

This article details the Tree Path Algorithm, which finds the first mutation step to convert a source syntax tree into a target tree.

Overcoming Ceiling Performance: Using Complexity Filtering for Harder Inverse Graphics Benchmarks

26 Sept 2025

This article addresses the challenge of creating sufficiently complex test sets for inverse graphics by using complexity filtering.

From Program to Sketch: Modeling Non-Deterministic Observations in Code Generation

26 Sept 2025

The Grammar of Code Generation: Detailed CFG Specifications for Graphics Languages

26 Sept 2025

This article provides the complete context-free grammar (CFG) specifications for the domain-specific graphics languages used in this research.

Controlling Program Length in Tree Diffusion: A Modified Mutation Sampling Algorithm

26 Sept 2025

This article provides a detailed breakdown of the mutation sampling algorithm for Tree-Diffusion, focusing on how to generate syntactically valid replacements.

The Future of Code Generation: Tree-Diffusion, Limitations, and Research Directions

26 Sept 2025

This conclusion summarizes a novel approach to program synthesis using a neural diffusion model operating on syntax trees.

The Importance of a Feedback Loop: An Ablation Study on Neural Code Generation

25 Sept 2025

This article presents an ablation study on the Tree-Diffusion model to evaluate the impact of its key design decisions.

The New Standard for Program Synthesis: How Tree-Diffusion Outperforms CSGNet and REPL Flow

25 Sept 2025

This article compares a novel denoising diffusion model for code generation (Tree-Diffusion) against two baseline methods, CSGNet and REPL Flow.