# PLAD: Learning to Infer Shape Programs with Pseudo-Labels and Approximate Distributions

arXiv: https://arxiv.org/abs/2011.13045 · HTML: https://mlsanity.com/paper/2011.13045/

## Key facts

- Cited by 19 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 25 Nov 2020.

## Built on by (most cited first, 19 of 19)

- [SceneScript: Reconstructing Scenes With An Autoregressive Structured Language Model](https://mlsanity.com/paper/2403.13064/) (2024)
- [Real2Code: Reconstruct Articulated Objects via Code Generation](https://mlsanity.com/paper/2406.08474/) (2024)
- [D2CSG: Unsupervised Learning of Compact CSG Trees with Dual Complements and Dropouts](https://mlsanity.com/paper/2301.11497/) (2023)
- [Can Large Language Models Understand Symbolic Graphics Programs?](https://mlsanity.com/paper/2408.08313/) (2024)
- [ShapeCoder: Discovering Abstractions for Visual Programs from Unstructured Primitives](https://mlsanity.com/paper/2305.05661/) (2023)
- [MeshCoder: LLM-Powered Structured Mesh Code Generation from Point Clouds](https://mlsanity.com/paper/2508.14879/) (2025)
- [Design2GarmentCode: Turning Design Concepts to Tangible Garments Through Program Synthesis](https://mlsanity.com/paper/2412.08603/) (2024)
- [From Perception to Programs: Regularize, Overparameterize, and Amortize](https://mlsanity.com/paper/2206.05922/) (2022)
- [BrepGPT: Autoregressive B-rep Generation with Voronoi Half-Patch](https://mlsanity.com/paper/2511.22171/) (2025)
- [Improving Unsupervised Visual Program Inference with Code Rewriting Families](https://mlsanity.com/paper/2309.14972/) (2023)
- [Learning to Edit Visual Programs with Self-Supervision](https://mlsanity.com/paper/2406.02383/) (2024)
- [DPA-Net: Structured 3D Abstraction from Sparse Views via Differentiable Primitive Assembly](https://mlsanity.com/paper/2404.00875/) (2024)
- [GIFT: Bootstrapping Image-to-CAD Program Synthesis via Geometric Feedback](https://mlsanity.com/paper/2603.27448/) (2026)
- [Tree-Structured Shading Decomposition](https://mlsanity.com/paper/2309.07122/) (2023)
- [Learning to Infer Generative Template Programs for Visual Concepts](https://mlsanity.com/paper/2403.15476/) (2024)
- [Learning to Place Objects with Programs and Iterative Self Training](https://mlsanity.com/paper/2503.04496/) (2025)
- [PS-CAD: Local Geometry Guidance via Prompting and Selection for CAD Reconstruction](https://mlsanity.com/paper/2405.15188/) (2024)
- [PLLM: Pseudo-Labeling Large Language Models for CAD Program Synthesis](https://mlsanity.com/paper/2602.12561/) (2026)
- [Pattern Analogies: Learning to Perform Programmatic Image Edits by Analogy](https://mlsanity.com/paper/2412.12463/) (2024)

## Abstract (arXiv)

Inferring programs which generate 2D and 3D shapes is important for reverse engineering, editing, and more. Training models to perform this task is complicated because paired (shape, program) data is not readily available for many domains, making exact supervised learning infeasible. However, it is possible to get paired data by compromising the accuracy of either the assigned program labels or the shape distribution. Wake-sleep methods use samples from a generative model of shape programs to approximate the distribution of real shapes. In self-training, shapes are passed through a recognition model, which predicts programs that are treated as "pseudo-labels" for those shapes. Related to these approaches, we introduce a novel self-training variant unique to program inference, where program pseudo-labels are paired with their executed output shapes, avoiding label mismatch at the cost of an approximate shape distribution. We propose to group these regimes under a single conceptual framework, where training is performed with maximum likelihood updates sourced from either Pseudo-Labels or an Approximate Distribution (PLAD). We evaluate these techniques on multiple 2D and 3D shape program inference domains. Compared with policy gradient reinforcement learning, we show that PLAD techniques infer more accurate shape programs and converge significantly faster. Finally, we propose to combine updates from different PLAD methods within the training of a single model, and find that this approach outperforms any individual technique.