Open Problems

3D Shape Modeling & Surface Reconstruction

Cross-Category and Vocabulary Generalization Benchmark for Text-to-3D Shape Models

UnclassifiedOpen
Weak candidate · 1/4 runs3 papers report this0% from 2025+

Generated automatically from the limitations stated in 3 papers (CVPR), listed under Evidence. It is not a paper, and it does not come from papers submitted to CSPaper.

The problem

Current text-driven 3D shape generation and modeling methods are predominantly trained and benchmarked strictly on Text2Shape, which is limited to chairs and tables (~11.5k shapes and a ~3.6k-word vocabulary). Consequently, it is unknown whether existing architectural designs, text encodings, and shape priors maintain generative fidelity when exposed to out-of-domain object categories or open-vocabulary text prompts. Without evaluating existing methods across broader category and vocabulary distributions, the field cannot determine if current performance gains reflect generalizable shape modeling or dataset-specific memorization.

Why it matters

Establishes the empirical boundaries of current text-conditioned 3D shape generation methods and provides a standardized protocol for evaluating cross-category robustness.

Ways to approach it

Prior-work checks are free with an account. Results someone already ran are shown to everyone.

  1. 1

    Systematic multi-category benchmark: Evaluate existing open-source Text2Shape models zero-shot and with fine-tuning on diverse ShapeNet classes (e.g., cars, airplanes, vessels) and measure text-shape retrieval accuracy, Chamfer Distance, and F-score across unseen classes.

  2. 2

    Open-vocabulary prompt stress-testing: Construct an evaluation suite of synthetic and real-world natural language descriptions spanning compositional and out-of-vocabulary terms to quantify failure modes in semantic binding and geometry generation.

  3. 3

    Patch- and component-level transfer evaluation: Assess whether localized geometric priors transfer across dissimilar categories when conditioned on shared compositional descriptions (e.g., legs, flat surfaces, thin supports).

Have a different approach?

Describe how you would tackle this problem and we'll look for papers that already do it. Free; your text stays private.

Free · 3 checks per day

Why it might fail

The problem could be dissolved if recent large-scale open-vocabulary 3D generative frameworks (e.g., trained on Objaverse) render chair/table-constrained Text2Shape baselines entirely obsolete for research benchmarking.

Evidence

Each paper's own statement of the limitation, verbatim.

Nearest existing work

Related open problems

3D Shape Modeling & Surface Reconstruction

Barrier to removeOpen

Reducing the Computational and Fidelity Cost of Distilling 2D Diffusion Priors into Explicit 3D Representations

Every pipeline that extracts 3D from pretrained 2D diffusion models currently pays a massive per-object tax: hundreds of rendered views, hundreds of diffusion evaluations, per-scene hyperparameter hand-tuning, and per-identity training — and still inherits the 2D prior's view-inconsistency and bias toward common categories. Nobody can train a single shared 3D generative model end-to-end on a shape dataset at acceptable cost, so 3D diffusion remains per-instance rather than dataset-scale. This blocks both generalization guarantees and practical deployment.

Possible candidate · 3/5 runs9 papers report this40% from 2025+
Generated automatically, not curated by hand. Automated prior-work checks catch about a third of existing work, so treat this problem as a lead to investigate.