AI Paper Atlas

Research map / Benchmarks / OK-VQA

OK-VQA leaderboard

71 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on OK-VQA. Best accuracy: 91.9 (StaR-KVQA (Gemma), CVPR 2026).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

OK-VQA · accuracy

60 papers, each with its best reported number for its own method; higher is better.

#accuracy ↑MethodPaperTopic
191.9StaR-KVQA (Gemma)
Gemma-3-12B
StaR-KVQA: Structured Reasoning Traces for Implicit-Knowledge Visual Question Answering
CVPR 2026 · Table 1
Visual question answering
285.6ProMSA
Qwen3-VL-8B
ProMSA:Progressive Multimodal Search Agents for Knowledge-Based Visual Question Answering
ECCV 2026 · Table 7
Visual question answering
376.5MATA (Domain-Specific)
InternVL2.5 (8B) + Qwen3 (4B) controller
MATA: A Trainable Hierarchical Automaton System for Multi-Agent Visual Reasoning
ICLR 2026 · Table 2
LLM agents
467.8DRESS
Vicuna-13B-v1.5
DRESS: Instructing Large Vision-Language Models to Align and Interact with Humans via Natural Language Feedback
CVPR 2024 · Table 5
Multimodal LLMs
566.2OURO-7B
Qwen2-VL-7B-Instruct
OURO: A Self-Bootstrapped Framework for Enhancing Multimodal Scene Understanding
ICCV 2025 · Table 1
Multimodal LLMs
665.9DDT-LLaMA
LLaMA3-8B
Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens
CVPR 2025 · Table 3
Multimodal LLMs
765.7Prophet++ (mPLUG)Prophet: Prompting Large Language Models With Complementary Answer Heuristics for Knowledge-Based Visual Question Answering
TPAMI 2025 · Table 3
Visual question answering
865.2LLaVA-1.5 + Q&A Prompts
LLaVA-1.5
Q&A Prompts: Discovering Rich Visual Clues through Mining Question-Answer Prompts for VQA requiring Diverse World Knowledge
ECCV 2024 · Table 2
Visual question answering
964.8Emu2-ChatGenerative Multimodal Models are In-Context Learners
CVPR 2024 · Table 2
Visual question answering
1063.8InstructBLIP w/ VIGC
InstructBLIP
VIGC: Visual Instruction Generation and Correction
AAAI 2024 · Table 4
Multimodal LLMs
50 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

OK-VQA · accuracy (val)

11 papers, each with its best reported number for its own method; higher is better.

#accuracy ↑MethodPaperTopic
168.2QACap + Claude 3.5
Claude 3.5
Separation of Powers: On Segregating Knowledge from Observation in LLM-enabled Knowledge-based Visual Question Answering
CVPR 2025 · Table 1
Visual question answering
266.1PaLI-X (55B)On Scaling Up a Multilingual Vision and Language Model
CVPR 2024 · Table 1
Visual question answering
366.1PaLM-E-562B
PaLM-E-562B
PaLM-E: An Embodied Multimodal Language Model
ICML 2023 · Table 4
Robot action and navigation
464.5PaLI-17B
mT5-XXL + ViT-e
PaLI: A Jointly-Scaled Multilingual Language-Image Model
ICLR 2023 · Table 2
Human and embodied modeling
562.4Omni-SMoLA PaLI-X FT
PaLI-X FT
Omni-SMoLA: Boosting Generalist Multimodal Models with Soft Mixture of Low-rank Experts
CVPR 2024 · Table 1
Multimodal LLMs
662.4Cola-FT
BLIP+OFA, FLAN-T5 11B
Large Language Models are Visual Reasoning Coordinators
NeurIPS 2023 · Table 1
Visual question answering
762Qwen-VL-7B + MTV
Qwen-VL-7B
Multimodal Task Vectors Enable Many-Shot Multimodal In-Context Learning
NeurIPS 2024 · Table 1
Human and embodied modeling
861.5LLaVA-HR-X
LLaVA-1.5
Feast Your Eyes: Mixture-of-Resolution Adaptation for Multimodal Large Language Models
ICLR 2025 · Table: Comparison with existin
Multimodal LLMs
960.4PromptCap + GPT-3
GPT-3 code-davinci-002 (175B)
PromptCap: Prompt-Guided Image Captioning for VQA with GPT-3
ICCV 2023 · Table 1
Visual question answering
1057.33LION-12B
FlanT5-XXL (11B)
LION: Empowering Multimodal Large Language Model with Dual-Level Visual Knowledge
CVPR 2024 · Table 1
Multimodal LLMs
1 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.