SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2503.22420 · CVPR · 2025

Unveiling the Mist over 3D Vision-Language Understanding: Object-centric Evaluation with Chain-of-Analysis

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 8 functions out of this paper's own repositories and ran 6 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
beacon-3d/beacon-3d canonical 6 of 8
FunctionStatusWhere it lives
build_results_mapping Ran beacon-3d/beacon-3d/evaluate_grounding.py
code served (permissive licence) · get_code("f391097e188013e2")
extract_number Ran beacon-3d/beacon-3d/utils.py
code served (permissive licence) · get_code("64da74993d3cde5c")
extract_pred Ran beacon-3d/beacon-3d/evaluate_grounding.py
code served (permissive licence) · get_code("d2b2260ca60666e1")
extract_pred Ran beacon-3d/beacon-3d/evaluate_qa.py
code served (permissive licence) · get_code("0504f9ada11d9af0")
is_binary_question Ran beacon-3d/beacon-3d/utils.py
code served (permissive licence) · get_code("9e1f3f2f2aa81886")
process_to_metadata Ran beacon-3d/beacon-3d/evaluate_qa.py
code served (permissive licence) · get_code("e42d8f07424c10ae")
call_openai_api_azure Not yet run beacon-3d/beacon-3d/utils.py
code served (permissive licence) · get_code("f0bdb1c7f27f1a6b")
process_to_metadata Not yet run beacon-3d/beacon-3d/evaluate_grounding.py
code served (permissive licence) · get_code("f2e6cf9e418a1c7a")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Figure 1. An overview of BEACON3D, a novel benchmark for 3D grounding and question answering (QA) tasks. BEACON3D features an object-centric evaluation framework, with Grounding-Chains (G-Chains) and Grounding-QA-Chains (GQA-Chains) for each object. The evaluation adopts object-centric metrics to ensure robustness and utilizes chain-of-analysis for studies in task coherence. We also involve the study of various knowledge types such as class, appearance ("App."), spatial ("Spa."), and geometry ("Geo.").

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2503.22420")
get_code_for_paper("2503.22420")
have("2503.22420")

Connect an agent — have() is free.