[arXiv-2025] Soft Thinking: Unlocking the Reasoning Potential of LLMs in Continuous Concept Space

[arXiv-2025] Hybrid Latent Reasoning via Reinforcement Learning