ExpertVerse: A General-Purpose Benchmark for Expert-Level Reasoning in Knowledge-Intensive Visual Synthesis
2026-07-21 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors created ExpertVerse, a detailed test set to check how well image-generating models use deep knowledge to make images, not just simple edits. Their test examines 9 types of thinking skills across 8 expert fields, with over 1,600 examples and a much larger 100K dataset including explanations. They also made KnowThinker, a model that learns to think step-by-step and improve instructions using a new training method called Bootstrapped Pareto Policy Optimization. Their experiments show current models struggle with complex reasoning, emphasizing the need for better knowledge-based image generation tests.
multimodal generative modelsknowledge-intensive generationcognitive capabilitiesvisual reasoningreinforcement learningBootstrapped Pareto Policy Optimizationreward rectificationimage generation benchmarkscross-modal learning
Authors
Yuan Wang, Yongchao Du, Mengting Chen, Jinsong Lan, Xuetao Feng, Xiaoyong Zhu
Abstract
Recent advances in multimodal generative models have enabled instruction-based image generation to move beyond semantic manipulation to knowledge-driven visual reasoning. However, these methods focus on explicit commonsense reasoning, shallow causal understanding, and direct knowledge recall, failing at knowledge-intensive generation. We develop \textbf{ExpertVerse}, a capability-centric benchmark to evaluate generative models via knowledge-intensive lens. ExpertVerse stratifies reasoning generation across an orthogonal taxonomy of \textit{9 cognitive capabilities} and \textit{8 expert disciplines}, yielding \textit{58 sub-disciplines}. We curate 1,611 expert-annotated instances covering single-image editing, multi-image composition, and text-to-image generation. We further develop an automated workflow to produce \textbf{ExpertVerse-100K}, a large-scale dataset with reasoning traces and knowledge-anchored rationale annotations. Based on this, we train \textbf{KnowThinker} with RL fine-tuning, a VLM reasoning engine with world knowledge that jointly generates thinking processes and refined instructions. Towards the cross-modal credit misalignment and multi-objective gradient conflicts in multi-reward optimization, we propose a tailored Bootstrapped Pareto Policy Optimization (BPPO), which synergizes Bootstrapping Reward Rectification (BRR) and Conflict-Aware Pareto Advantage Fusion (CPAF). Extensive results of both open-source and proprietary models exposes critical reasoning deficits, highlighting imperative for knowledge-intensive benchmarks towards next-generation visual generation.