PLUME: Latent Reasoning Based Universal Multimodal Embedding

TL;DR

PLUME introduces a latent reasoning framework for universal multimodal embedding that replaces explicit chain-of-thought with continuous latent states, achieving faster inference and improved performance on complex multimodal tasks.

Contribution

It proposes a novel latent reasoning approach with a semantic-anchor-guided transition adapter and a progressive training curriculum, outperforming explicit-CoT methods in speed and accuracy.

Findings

Outperforms explicit-CoT UME baselines on MMEB-v2 benchmark.

Reduces reasoning steps from hundreds to fewer than 10 latent steps.

Achieves over 30x faster inference in retrieval tasks.

Abstract

Universal multimodal embedding (UME) maps heterogeneous inputs into a shared retrieval space with a single model. Recent approaches improve UME by generating explicit chain-of-thought (CoT) rationales before extracting embeddings, enabling multimodal large language models to better infer complex query intent. However, explicit CoT incurs substantial inference overhead and can compress rich multimodal evidence into a narrow textual bottleneck. We propose PLUME, a latent reasoning framework that advances UME by replacing verbalized CoT with a short autoregressive rollout of continuous latent states. To support diverse multimodal queries, PLUME further introduces a semantic-anchor-guided transition adapter that steers latent rollout along different reasoning trajectories under the same fixed computation budget. To stabilize training, PLUME adopts a progressive explicit-to-latent curriculum that uses verbalized reasoning only as a temporary training scaffold and gradually transfers this behavior into hidden-state computation, eliminating explicit CoT at inference. On the 78-task MMEB-v2 benchmark, PLUME outperforms strong explicit-CoT UME baselines while reducing reasoning from hundreds of generated tokens to fewer than 10 latent steps, delivering over 30x faster inference. PLUME is especially well suited to retrieval settings where relevant evidence is dense, structurally complex, and difficult to organize through verbalized intermediate rationales, such as video and visual document retrieval. These results show that structured latent computation can preserve the benefits of intermediate reasoning without the overhead of explicit rationale generation, providing a stronger and more efficient paradigm for practical retrieval systems.