# Multimodal Latent Reasoning via Predictive Embeddings

arXiv: https://arxiv.org/abs/2604.08065 · HTML: https://mlsanity.com/paper/2604.08065/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Multimodal reasoning, one of 17 directions in LLM reasoning.
- Builds on 18 of its 31 references that are in the map.
- First posted on arXiv 9 Apr 2026.

## Placed in

- LLM reasoning / [Multimodal reasoning](https://mlsanity.com/lr/multimodal-reasoning/)
- Multimodal LLMs / vision-language models / [VLM benchmarks and visual reasoning](https://mlsanity.com/mllm/vlm-benchmarks-and-visual-reasoning/)

## Builds on (most cited first, 18 of 18)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Qwen3-VL Technical Report](https://mlsanity.com/paper/2511.21631/) (2025)
- [V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning](https://mlsanity.com/paper/2506.09985/) (2025)
- [Training Large Language Models to Reason in a Continuous Latent Space](https://mlsanity.com/paper/2412.06769/) (2024)
- [V*: Guided Visual Search as a Core Mechanism in Multimodal LLMs](https://mlsanity.com/paper/2312.14135/) (2023)
- [Pixel Reasoner: Incentivizing Pixel-Space Reasoning with Curiosity-Driven Reinforcement Learning](https://mlsanity.com/paper/2505.15966/) (2025)
- [OpenThinkIMG: Learning to Think with Images via Visual Tool Reinforcement Learning](https://mlsanity.com/paper/2505.08617/) (2025)
- [Can MLLMs Reason in Multimodality? EMMA: An Enhanced MultiModal ReAsoning Benchmark](https://mlsanity.com/paper/2501.05444/) (2025)
- [Machine Mental Imagery: Empower Multimodal Reasoning with Latent Visual Tokens](https://mlsanity.com/paper/2506.17218/) (2025)
- [Latent Visual Reasoning](https://mlsanity.com/paper/2509.24251/) (2025)
- [Chain-of-Visual-Thought: Teaching VLMs to See and Think Better with Continuous Visual Tokens](https://mlsanity.com/paper/2511.19418/) (2025)
- [VL-JEPA: Joint Embedding Predictive Architecture for Vision-language](https://mlsanity.com/paper/2512.10942/) (2025)
- [ThinkMorph: Emergent Properties in Multimodal Interleaved Chain-of-Thought Reasoning](https://mlsanity.com/paper/2510.27492/) (2025)
- [LLM-JEPA: Large Language Models Meet Joint Embedding Predictive Architectures](https://mlsanity.com/paper/2509.14252/) (2025)
- [VisualToolAgent (VisTA): A Reinforcement Learning Framework for Visual Tool Selection](https://mlsanity.com/paper/2505.20289/) (2025)
- [Next-Latent Prediction Transformers Learn Compact World Models](https://mlsanity.com/paper/2511.05963/) (2025)
- [v1: Learning to Point Visual Tokens for Multimodal Grounded Reasoning](https://mlsanity.com/paper/2505.18842/) (2025)
- [Optimizing Agentic Language Model Inference via Speculative Tool Calls](https://mlsanity.com/paper/2512.15834/) (2025)

## Built on by (most cited first, 4 of 4)

- [LUT: Latent Utility Training for Visual Reasoning](https://mlsanity.com/paper/2608.00743/) (2026)
- [Reason Through the Latent! Making Latent Visual Reasoning Necessary](https://mlsanity.com/paper/2609.06746/) (2026)
- [DeepLatent: Think with Images via Parallel Latent Visual Reasoning](https://mlsanity.com/paper/2606.00562/) (2026)
- [GLaQ: Grounding Latent Queries in Visual Evidence for Multimodal Reasoning](https://mlsanity.com/paper/2608.15517/) (2026)

## Abstract (arXiv)

Tool-augmented multimodal reasoning enables visual language models (VLMs) to improve perception by interacting with external tools (e.g., cropping, depth estimation). However, such approaches incur substantial inference overhead, require specialized supervision, and are prone to erroneous tool calls. We propose Pearl (Predictive Embedding Alignment for Reasoning in Latent space), a JEPA-inspired framework that learns from expert tool-use trajectories entirely in the latent space, eliminating the need for explicit tool invocation at inference time. Unlike reconstruction-based latent reasoning methods, which autoregressively generate latent tokens and suffer from training-inference mismatch and limited support for multi-step tool use, Pearl directly learns predictive embeddings from multimodal trajectories while preserving the standard vision-language generation pipeline: it is model-agnostic, simple to train, and naturally supports trajectories with multiple tool calls. Experiments across multiple perception benchmarks show that Pearl matches or outperforms standard supervised fine-tuning and reconstruction-based latent reasoning approaches. Furthermore, we provide empirical evidence that reconstruction-based methods primarily learn embeddings rather than image edits in latent space, motivating predictive embedding learning as a more principled alternative.