Can MLLMs Decode the Creative Leap? Introducing C4 for Cross-Concept Understanding

2026-08-06

AI Summary

This work addresses the challenge of evaluating cross-conceptual understanding in multimodal large language models (MLLMs) within open-ended creative tasks. To this end, the authors propose the C4 framework, which formalizes cross-conceptual creativity as a measurable cognitive task by leveraging Chinese idioms as conceptual carriers. They construct C4-Eval, a benchmark comprising both synthetic and human-authored samples, accompanied by structured difficulty metrics, explicit reference answers, and human-validated bridging paths. Through techniques including cross-conceptual network modeling, candidate constraint enforcement, and prompt-based interventions, the study systematically assesses MLLMs’ creative decoding capabilities. Experimental results reveal that even the strongest closed-source models achieve only around 50% accuracy on the primary task, while open-source counterparts lag significantly; candidate constraints yield notable performance gains, whereas bridging prompts offer limited improvement, underscoring fundamental limitations in current models’ capacity for creative conceptual synthesis.

Abstract

Creative capabilities of MLLMs matter in design, communication, education, and human--AI collaboration, yet remain difficult to evaluate because explicit targets and reward signals are scarce compared with accuracy-oriented tasks. Cross-concept understanding is a core cognitive capacity underlying receptive creativity. It enables a perceiver to recover intended meaning from non-obvious but meaningful conceptual relations. We operationalize item construction as cross-concept encoding and model inference as cross-concept decoding. We introduce C4, a cognition-inspired evaluation framework for Chengyu (Chinese idiom)-based Cross-Concept Creativity. Its encoding component maps target slots to imageable substitute concepts along bridge paths in a manually annotated and third-party-reviewed cross-concept network, enabling batch generation with explicit structure, difficulty indexed by bridge count and depth, and exact answers. Using this framework, we instantiate the C4 Evaluation Set (C4-Eval), comprising 184 synthetic items and 37 human-created cross-concept chengyu figures collected from online sources. We manually construct and review cross-concept relations, bridge paths, and reasoning processes for the collected figures. Each C4-Eval item is instantiated in five task settings, yielding 884 primary answer-recovery cases. Across ten evaluated MLLMs, the strongest closed models reach 50.7% and 48.0% primary accuracy, while open-source models remain substantially lower. Candidate constraints improve accuracy sharply, but bridge hints and explanation requests provide only modest gains. These results expose a substantial gap in how current MLLMs decode creatively encoded meaning through cross-concept relations. The code is in the supplementary material.

Problem

• cross-concept understanding
• creative decoding
• MLLMs
• receptive creativity
• Chengyu

Innovation

• cross-concept understanding
• creative decoding
• C4 framework
• chengyu-based creativity
• multimodal LLM evaluation