ArtMine: Discovering and Formalizing Artistic Processes

2026-07-11 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Computational Art History · Depth: Expert, extended

Summary

ArtMine is a framework designed to discover and formalize artistic processes from fragmented historical evidence, moving beyond generative AI's typical focus on finished artifacts. It synthesizes heterogeneous sources like archival records and preparatory studies into a structured repository with 11 dimensions. A Peircean abductive agent, powered by Qwen2.5-VL, infers evidence-grounded production steps, which are then converted into a compositional graph and rendering prompt. This process is optimized through self-reflection, comparing generated and reference artworks using multi-reward feedback. A preliminary case study across 10 artworks, including canonical pieces like "Mona Lisa" and "The Scream," demonstrates ArtMine's ability to reconstruct coherent, interpretable, and auditable artistic workflows, outperforming baselines (CoT, CoT-SC, ToT, Self-Refine) with image generation handled by FLUX.1-dev.

Key takeaway

For AI Scientists and Research Scientists developing generative models for creative or historical domains, ArtMine offers a critical shift from artifact-centric generation to process-centered understanding. You should consider integrating structured evidence repositories, abductive reasoning, and self-reflection mechanisms into your systems. This approach enables the reconstruction of interpretable, auditable creative workflows, fostering more meaningful human-AI co-creativity and deeper insights into cultural production, rather than merely synthesizing final outputs.

Key insights

ArtMine formalizes artistic processes from historical evidence using abductive reasoning and self-reflection, shifting AI from artifact generation to process modeling.

Principles

• Abductive inference can reconstruct unobserved creative trajectories from fragmented evidence.
• Structured evidence repositories with reliability tags enhance reasoning and conflict resolution.
• Policy learning via visual feedback loops iteratively refines generative process models.

Method

ArtMine employs a Deep Research agent (MiroThinker 1.7 mini) for structured evidence, a Qwen2.5-VL-based abductive agent for step inference, and multi-reward self-reflection for policy optimization and visual refinement.

In practice

• Implement multi-agent systems for evidence synthesis and process inference.
• Utilize Peircean abductive reasoning for explaining latent creative actions.
• Integrate visual feedback to iteratively improve generative process outputs.

Topics

• ArtMine
• Artistic Process Discovery
• Abductive Reasoning
• Generative AI
• Human-AI Co-Creativity
• Computational Art History

Best for: AI Scientist, Research Scientist

Related on AIssential

• ArtMine: Discovering and Formalizing Artistic Processes — ArtMine formalizes artistic processes from fragmented historical data, enabling process-centered human-AI co-creativity beyond artifact generation.
• ArtMine: Discovering and Formalizing Artistic Processes — ArtMine formalizes artistic processes from fragmented historical data using AI for co-creativity, moving beyond artifact generation.
• Back to Back with a Copy: A Computational Analysis of AI-Generated Visual Contemporary Art Pastiches — Artistic style evaluation requires a multidimensional approach, as AI pastiches show a "compositional gap" between texture and semantics.
• Unboxing Diffusion Models for the Arts: Interactive Model Bending and Practice-Based Explainability — Making diffusion model internals visible and manipulable empowers artists to creatively inspect, modify, and debug outputs.
• Visual Retrieval-Augmented Generation for Silhouette-Guided Animal Art — Visual-RAG computationally interprets ambiguous shapes, generating animal art from natural silhouettes using retrieval and diffusion models.
• Creo: From One-Shot Image Generation to Progressive, Co-Creative Ideation — Multi-stage T2I generation with intermediate control enhances user agency, creativity, and output diversity.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.