Research map / Vision-language and multimodal / Multimodal LLMs

Composition-Grounded Instruction Synthesis for Visual Reasoning

ICLR 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes COGS, a composition-grounded method that synthesizes reasoning instruction data for artificial image domains (charts, rendered documents, webpages) to improve MLLM visual reasoning without human annotations.

Problem

MLLMs remain limited in reasoning on artificial image domains such as charts, rendered documents, and webpages, where large-scale human-annotated reasoning datasets are lacking.

Approach

Synthesizes reasoning instructions grounded in compositions of visual elements, rather than relying on human annotation; the abstract is truncated so the specific composition mechanism is not stated.

Closest papers by meaning

• SLVR: Structured Latent Visual Reasoning via Human-like Reasoning Flows (NeurIPS 2026 · similarity 0.90)
• Visual Graph Arena: Evaluating Visual Conceptualization of Vision and Multimodal Large Language Models (ICML 2025 · similarity 0.90)
• ChartMuseum: Testing Visual Reasoning Capabilities of Large Vision-Language Models (NeurIPS 2025 · similarity 0.90)
• Long Grounded Thoughts: Synthesizing Visual Problems and Reasoning Chains at Scale (ICML 2026 · similarity 0.90)
• VGR: Visual Grounded Reasoning (ICLR 2026 · similarity 0.90)
• AStar: Boosting Multimodal Reasoning with Automated Structured Thinking (AAAI 2026 · similarity 0.90)