Machine Mental Imagery: Empower Multimodal Reasoning with Latent Visual Tokens

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Yang, Zeyuan, Yu, Xueyang, Chen, Delin, Shen, Maohao, Gan, Chuang
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912443040530432
author Yang, Zeyuan
Yu, Xueyang
Chen, Delin
Shen, Maohao
Gan, Chuang
author_facet Yang, Zeyuan
Yu, Xueyang
Chen, Delin
Shen, Maohao
Gan, Chuang
contents Vision-language models (VLMs) excel at multimodal understanding, yet their text-only decoding forces them to verbalize visual reasoning, limiting performance on tasks that demand visual imagination. Recent attempts train VLMs to render explicit images, but the heavy image-generation pre-training often hinders the reasoning ability. Inspired by the way humans reason with mental imagery-the internal construction and manipulation of visual cues-we investigate whether VLMs can reason through interleaved multimodal trajectories without producing explicit images. To this end, we present a Machine Mental Imagery framework, dubbed as Mirage, which augments VLM decoding with latent visual tokens alongside ordinary text. Concretely, whenever the model chooses to ``think visually'', it recasts its hidden states as next tokens, thereby continuing a multimodal trajectory without generating pixel-level images. Begin by supervising the latent tokens through distillation from ground-truth image embeddings, we then switch to text-only supervision to make the latent trajectory align tightly with the task objective. A subsequent reinforcement learning stage further enhances the multimodal reasoning capability. Experiments on diverse benchmarks demonstrate that Mirage unlocks stronger multimodal reasoning without explicit image generation.
format Preprint
id arxiv_https___arxiv_org_abs_2506_17218
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Machine Mental Imagery: Empower Multimodal Reasoning with Latent Visual Tokens
Yang, Zeyuan
Yu, Xueyang
Chen, Delin
Shen, Maohao
Gan, Chuang
Computer Vision and Pattern Recognition
Artificial Intelligence
Vision-language models (VLMs) excel at multimodal understanding, yet their text-only decoding forces them to verbalize visual reasoning, limiting performance on tasks that demand visual imagination. Recent attempts train VLMs to render explicit images, but the heavy image-generation pre-training often hinders the reasoning ability. Inspired by the way humans reason with mental imagery-the internal construction and manipulation of visual cues-we investigate whether VLMs can reason through interleaved multimodal trajectories without producing explicit images. To this end, we present a Machine Mental Imagery framework, dubbed as Mirage, which augments VLM decoding with latent visual tokens alongside ordinary text. Concretely, whenever the model chooses to ``think visually'', it recasts its hidden states as next tokens, thereby continuing a multimodal trajectory without generating pixel-level images. Begin by supervising the latent tokens through distillation from ground-truth image embeddings, we then switch to text-only supervision to make the latent trajectory align tightly with the task objective. A subsequent reinforcement learning stage further enhances the multimodal reasoning capability. Experiments on diverse benchmarks demonstrate that Mirage unlocks stronger multimodal reasoning without explicit image generation.
title Machine Mental Imagery: Empower Multimodal Reasoning with Latent Visual Tokens
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2506.17218