OWLViz: An Open-World Benchmark for Visual Question Answering

ACL ARR 2025 July Submission744 Authors

Abstract: We present a challenging benchmark for the Open WorLd VISual (OWLViz) question answering benchmark. \name~ presents short queries that require integrating multiple capabilities, including common-sense knowledge, visual understanding, web exploration, and specialized tool usage. While humans achieve 69.2% accuracy on these intuitive tasks, even state-of-the-art VLMs struggle, with the best model, Gemini 2.0, achieving only 26.6% accuracy. Current agentic VLMs, which rely on limited vision and vision-language models as tools, perform even worse. This performance gap reveals significant limitations in multimodal systems' ability to select appropriate tools and execute complex reasoning sequences, establishing new directions for advancing practical AI research.

Paper Type: Long

Research Area: Resources and Evaluation

Research Area Keywords: Open World, Tool Use, VQA

Contribution Types: Data resources

Languages Studied: English

Previous URL: https://openreview.net/forum?id=oxgRh77rxv

Reassignment Request Area Chair: No, I want the same area chair from our previous submission (subject to their availability).

Reassignment Request Reviewers: No, I want the same set of reviewers from our previous submission (subject to their availability)

A1 Limitations Section: This paper has a limitations section.

A2 Potential Risks: No

A2 Elaboration: This paper offers a small dataset for evaluating system. We did not release the golden data to prevent data contamination. We don't foresee critical potential risks.

B Use Or Create Scientific Artifacts: Yes

B1 Cite Creators Of Artifacts: Yes

B1 Elaboration: Section 1, 2, 3.5, 4

B2 Discuss The License For Artifacts: Yes

B2 Elaboration: Ethical consideration

B3 Artifact Use Consistent With Intended Use: Yes

B3 Elaboration: Ethical consideration

B4 Data Contains Personally Identifying Info Or Offensive Content: No

B4 Elaboration: The dataset does not include personal information.

B5 Documentation Of Artifacts: Yes

B5 Elaboration: Section 3

B6 Statistics For Data: Yes

B6 Elaboration: Section 3

C Computational Experiments: Yes

C1 Model Size And Budget: Yes

C1 Elaboration: Appendix B (Experiment Details)

C2 Experimental Setup And Hyperparameters: Yes

C2 Elaboration: Appendix B (Experiment Details)

C3 Descriptive Statistics: No

C3 Elaboration: The results were computed based on a single run

C4 Parameters For Packages: N/A

D Human Subjects Including Annotators: No

D1 Instructions Given To Participants: No

D1 Elaboration: We do not use crowd workers.

D2 Recruitment And Payment: N/A

D2 Elaboration: We do not use crowd workers.

D3 Data Consent: Yes

D3 Elaboration: We discuss this for evaluation, no golden data is released.

D4 Ethics Review Board Approval: N/A

D5 Characteristics Of Annotators: No

D5 Elaboration: The authors are annotators.

E Ai Assistants In Research Or Writing: No

E1 Information About Use Of Ai Assistants: No

E1 Elaboration: We wrote it ourselves.

Author Submission Checklist: yes

Submission Number: 744