ATLAS: Agentic or Latent Visual Reasoning? One Word is Enough for Both

TL;DR

ATLAS introduces a unified framework using functional tokens for visual reasoning that combines agentic and latent approaches, achieving high performance without complex visual content generation.

Contribution

The paper proposes a novel token-based reasoning framework that merges agentic and latent methods, improving efficiency and generalization in visual reasoning tasks.

Findings

ATLAS outperforms existing methods on challenging benchmarks.

The framework maintains interpretability and compatibility with standard training procedures.

Latent-Anchored GRPO stabilizes training of functional tokens.

Abstract

Visual reasoning, often interleaved with intermediate visual states, has emerged as a promising direction in the field. A straightforward approach is to directly generate images via unified models during reasoning, but this is computationally expensive and architecturally non-trivial. Recent alternatives include agentic reasoning through code or tool calls, and latent reasoning with learnable hidden embeddings. However, agentic methods incur context-switching latency from external execution, while latent methods lack task generalization and are difficult to train with autoregressive parallelization. To combine their strengths while mitigating their limitations, we propose ATLAS, a framework in which a single discrete 'word', termed as a functional token, serves both as an agentic operation and a latent visual reasoning unit. Each functional token is associated with an internalized visual operation, yet requires no visual supervision and remains a standard token in the tokenizer vocabulary, which can be generated via next-token prediction. This design avoids verbose intermediate visual content generation, while preserving compatibility with the vanilla scalable SFT and RL training, without architectural or methodological modifications. To further address the sparsity of functional tokens during RL, we introduce Latent-Anchored GRPO (LA-GRPO), which stabilizes the training by anchoring functional tokens with a statically weighted auxiliary objective, providing stronger gradient updates. Extensive experiments and analyses demonstrate that ATLAS achieves superior performance on challenging benchmarks while maintaining clear interpretability. We hope ATLAS offers a new paradigm inspiring future visual reasoning research.