benchmark.darvinyi.com
← Back to Benchmarks
MultimodalActive

WorldBench

A 2,000-question multimodal reasoning benchmark built to maximize visual diversity, not just task diversity, across seven real-world visual domains.

Tasks2,000
Year2026
CreatorYida Yin, Harish Krishnakumar, Chung Peng Lee, Boya Zeng, Wenhao Chai, Shengbang Tong, Wenhu Chen, Hu Xu, Xingyu Fu, Gabriel Sarch, Aleksandra Korolova, Zhuang Liu
MetricMultiple-choice accuracy (%), reported per visual domain and as an overall average across all 2,000 questions.

What It Tests

WorldBench is a multimodal reasoning benchmark from researchers at Princeton University (Yida Yin, Harish Krishnakumar, Chung Peng Lee, Boya Zeng, Wenhao Chai, Shengbang Tong, Wenhu Chen, Hu Xu, Xingyu Fu, Gabriel Sarch, Aleksandra Korolova, Zhuang Liu) designed to test whether multimodal large language models (MLLMs) can reason reliably across the full breadth of visual settings they encounter in the real world, rather than just a wide range of task formats. The authors argue that prior multimodal benchmarks expand task types (VQA, captioning, chart reading, etc.) while sampling images narrowly from a small number of iconic, object-centric sources, leaving a gap in visual diversity that lets models look strong without actually generalizing.

To close that gap, the authors built a large taxonomy of roughly 2,000 visual concepts spanning seven domains (Living Things, Objects, Scenes, Digital World, Academics, Documents/Charts/Tables, and Agents), then curated high-quality, deliberately non-iconic images for each concept -- images that represent a concept in a realistic, cluttered, or unusual context rather than a clean textbook photo. Questions were then hand-written through a structured trial-and-error process specifically to expose cases where frontier MLLMs answer incorrectly despite the task being natural and intuitive for a human. The result is 2,000 curated multiple-choice questions, each paired with an image.

What differentiates WorldBench is its explicit focus on measuring and maximizing visual diversity itself: the authors validate this with embedding-based diversity metrics (effective rank and participation ratio across three vision encoders) and human pairwise (Bradley-Terry) diversity ratings, both showing WorldBench is more visually diverse than existing 'diverse' benchmarks. On the evaluation side, the headline trajectory is that the benchmark is far from saturated -- the strongest model evaluated (Gemini-3.1-Pro) reaches only 64.0% average accuracy, and no model exceeds 75% accuracy in any single domain, with several open-source models performing only marginally above chance on a 4-option multiple-choice task.

Task Anatomy

How a single task is structured.

InputA single image drawn from a taxonomy of ~2,000 visual concepts across 7 domains (Living Things, Objects, Scenes, Digital World, Academics, Documents/Charts/Tables, Agents), curated to be non-iconic (realistic/cluttered rather than clean object-centric shots), paired with a 4-option multiple-choice question (option_a-option_d) about the image.
OutputA single selected answer letter (A, B, C, or D).
EvaluationAutomated answer extraction/normalization from the model's response, compared against the ground-truth answer letter; the official harness (Hydra-configured, supports proprietary and open-source models) computes per-domain and overall accuracy.
MetricMultiple-choice accuracy (%), reported per visual domain and as an overall average across all 2,000 questions.

Example Tasks

2 real examples from the benchmark.

#1

Living Things domain (representative format, not a verbatim quoted item)

Problem / Input

A non-iconic photograph depicting a living-thing concept from the taxonomy in a realistic, uncontrolled setting (e.g., partially occluded, camouflaged, or shown in an atypical context) rather than a clean, centered reference photo, accompanied by a 4-option multiple-choice question probing a specific visually-grounded detail (e.g., identity, attribute, count, or relationship) that requires actually looking at the image rather than guessing from common priors.
AnswerOne of four labeled options (A-D); ground truth fixed during curation and cross-checked so the question is 'natural for humans but challenging for frontier MLLMs.'

Constructed from the benchmark's documented construction methodology and dataset schema (image, question, option_a-d, answer, category, domain fields); the paper/project page/GitHub/HuggingFace pages accessed did not expose the literal text of individual questions, so no specific question wording is quoted here.

#2

Documents/Charts/Tables (DCT) domain (representative format, not a verbatim quoted item)

Problem / Input

An image of a real-world document, chart, or table sampled to reflect the visual diversity of that concept category, paired with a 4-option multiple-choice question requiring the model to read or reason over the depicted information.
AnswerOne of four labeled options (A-D).

Same caveat as above: format is confirmed via the HuggingFace dataset card schema and GitHub README, but exact question text was not available in the sources reviewed.

Leaderboard Results

Model scores sorted by performance.

15 results

Sort:
#ModelScore
1
Gemini 3.1 Pro
64.0% avg accuracy
2
Gemini 3 Flash
61.8% avg accuracy
3
Qwen3.5-VL-Plus (Thinking)
59.3% avg accuracy
4
GPT-5.4 Thinking
58.2% avg accuracy (Thinking, high effort)
5
Qwen3.5-VL-27B
56.6% avg accuracy
6
Claude Opus 4.7
53.7% avg accuracy
7
Grok 4.2
53.3% avg accuracy
8
GPT-5.4 Thinking
53.0% avg accuracy (Thinking, low effort)
9
Qwen3.5-VL-35B-A3B
52.9% avg accuracy
10
Kimi K2.5
52.5% avg accuracy
11
Gemma 4 31B
49.7% avg accuracy
12
Qwen3.5-VL-Plus (Instruct)
48.7% avg accuracy
13
GLM-4.6V
42.5% avg accuracy
14
InternVL3.5
41.2% avg accuracy
15
Gemma 4 E4B
34.6% avg accuracy

V= Self-reported by the model's creator, not independently verified

Score Over Time

Performance progression across model generations.

Key Findings

  • The best-performing model overall, Gemini-3.1-Pro, reaches only 64.0% average accuracy across the 2,000 questions, and no model exceeds 75% accuracy in any single one of the 7 domains.

  • The best open-source/open-weight model, Qwen3.5-VL-27B, reaches 56.6% average accuracy, trailing the top proprietary model by about 7.4 points.

  • Several smaller open models (e.g., Gemma-4-E4B at 34.6%, InternVL-3.5 at 41.2%) perform only marginally above chance level on the 4-option multiple-choice format, revealing large gaps in robustness to visually diverse, non-iconic imagery.

  • Embedding-based diversity metrics (effective rank / participation ratio across three vision encoders) and human pairwise Bradley-Terry ratings from 12 raters both confirm WorldBench achieves higher visual diversity than prior 'diverse' multimodal benchmarks.

  • The authors explicitly frame the benchmark's headline result as 'WorldBench Remains Far From Saturated,' positioning visual (not just task) diversity as the harder, still-unsolved axis of multimodal evaluation.

Controversies & Caveats

Known limitations and criticisms.

WorldBench was only posted to arXiv in June 2026, roughly a month before this research was conducted; no independent third-party audits, contamination analyses, or critical replications were found in the sources reviewed, so current scores rest entirely on the authors' own evaluation harness.

Some leaderboard entries (e.g., GPT-5.4-Thinking at 'high' vs 'low' reasoning effort) show non-trivial score swings from inference-time settings alone, which could complicate cross-model comparisons if effort settings aren't standardized in future reporting.

Links