ATLAS: Functional Tokens Enable Dual Agentic and Latent Visual Reasoning Without Architecture Changes

arXiv·medium signal

ATLAS introduces 'functional tokens' — discrete words added to the standard tokenizer vocabulary that serve as both agentic operation triggers and latent visual reasoning units. This avoids the context-switching latency of tool-call-based agentic reasoning and the generalization issues of learned hidden embeddings. The paper introduces Latent-Anchored GRPO (LA-GRPO) for stable RL training, and achieves superior performance on visual reasoning benchmarks while maintaining compatibility with standard SFT and RL training pipelines with no architectural modifications required.