SpiralThinker: Latent Reasoning through an Iterative Process with Text-Latent Interleaving
This addresses the need for stable and systematic latent reasoning in AI, though it appears incremental as it builds on existing latent reasoning methods.
The paper tackled the problem of latent reasoning in large models by introducing SpiralThinker, a framework that uses iterative updates over latent representations with text-latent interleaving, achieving the best overall performance among latent reasoning approaches across mathematical, logical, and commonsense reasoning benchmarks.
Recent advances in large reasoning models have been driven by reinforcement learning and test-time scaling, accompanied by growing interest in latent rather than purely textual reasoning. However, existing latent reasoning methods lack mechanisms to ensure stable evolution of latent representations and a systematic way to interleave implicit and explicit reasoning. We introduce SpiralThinker, a unified framework that performs iterative updates over latent representations, enabling extended implicit reasoning without generating additional tokens. A progressive alignment objective combined with structured annotations maintains coherence between latent and textual reasoning. Across mathematical, logical, and commonsense reasoning tasks, SpiralThinker achieves the best overall performance among latent reasoning approaches, consistently surpassing previous methods across all benchmarks. Detailed analyses reveal that both iteration and alignment are indispensable, the numbers of latent tokens and iterations exhibit dataset-specific optima, and appropriate alignment proves critical for an effective iterative process. Overall, SpiralThinker bridges iterative computation and latent reasoning, demonstrating that aligned iterative updates can reliably steer reasoning in the latent space.