R-4B：通过双模式退火与强化学习激发多模态大模型中的通用自动思考
R-4B: Incentivizing General-Purpose Auto-Thinking in MLLMs via Bi-Mode Annealing and Reinforce Learning
为视觉语言模型引入「按输入复杂度自动决定是否展开逐步推理」的双模式机制，配合双 rollout 强化学习策略，可直接迁移到需要兼顾精度与推理成本的 VQA、图文推理等视觉任务部署中。
原文依据 an auto-thinking MLLM that dynamically determines whether to invoke the reasoning process based on input complexity
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论具备显式逐步推理能力的多模态大语言模型（MLLMs）在复杂任务上已取得强劲表现。然而，对于许多简单查询而言，这种推理并不必要，还会带来可观的计算开销。为解决这一低效问题，我们提出 R-4B，一个自动思考的多模态大语言模型，它根据输入复杂度动态决定是否调用推理过程。我们的核心思路是让单一模型同时具备思考与非思考两种能力，并训练其选择合适的模式。我们首先引入双模式退火，这是一种统一的训练范式，无需显式的复杂度标注即可构建出在推理密集型与直接作答两种设定下均胜任的模型。在此基础上，我们提出双模式策略优化（BPO），一种轻量级强化学习算法，采用双 rollout 机制：对每个输入，模型同时生成思考与非思考两种回答。这避免了模式崩溃，并仅使用简单的基于规则的奖励即可稳健地学习自适应推理策略。在 25 个基准上的大量实验表明，R-4B 在同等规模模型中达到最优性能。它持续超越 Qwen2.5-VL-7B，并在推理密集型任务上匹配或超过 Kimi-VL-A3B-Thinking-2506（16B）等更大模型，同时通过避免冗余推理降低了计算成本。我们的结果表明，自适应自动思考为更高效的多模态推理模型提供了一条有效且可扩展的路径。
英文原文摘要 1595 字符
Multimodal Large Language Models (MLLMs) with explicit step-by-step reasoning have achieved strong performance on complex tasks. However, such reasoning is unnecessary for many simple queries and introduces substantial computational overhead. To address this inefficiency, we present R-4B, an auto-thinking MLLM that dynamically determines whether to invoke the reasoning process based on input complexity.Our key idea is to equip a single model with both thinking and non-thinking capabilities and train it to select the appropriate mode. We first introduce bi-mode annealing, a unified training paradigm that constructs a model competent in both reasoning-intensive and direct-answer settings without requiring explicit complexity annotations. Building on this foundation, we propose Bi-mode Policy Optimization (BPO), a lightweight reinforcement learning algorithm that employs a dual-rollout mechanism: for each input, the model generates both thinking and non-thinking responses. This prevents mode collapse and enables robust learning of an adaptive reasoning policy using only simple, rule-based rewards.Extensive experiments across 25 benchmarks show that R-4B achieves state-of-the-art performance among models of similar scale. It consistently surpasses Qwen2.5-VL-7B and matches or exceeds larger models such as Kimi-VL-A3B-Thinking-2506 (16B) on reasoning-intensive tasks, while reducing computational cost by avoiding redundant reasoning. Our results demonstrate that adaptive auto-thinking offers an effective and scalable pathway toward more efficient multimodal reasoning models.