AI 中文总结
针对交互式验证码对计算机使用智能体的挑战，提出大规模细粒度数据集CaptchaArena，训练单一9B策略CaptchaAgent，通过监督微调和强化学习提升求解性能。
AI 中文摘要
交互式验证码对计算机使用智能体仍然具有挑战性，而现有数据集在类型覆盖、交互保真度和轨迹监督之间存在权衡。为解决这些不足，我们提出了CaptchaArena，这是首个用于交互式验证码求解的大规模、细粒度训练数据集。它包含20种验证码类型和5种交互模式下的5万道谜题，每个解决方案都通过执行验证。CaptchaArena提供了5万条截图-动作轨迹，其中4.6万条带有逐步推理注释。它还包括针对不规则目标的细粒度像素掩码注释。利用CaptchaArena，我们训练了CaptchaAgent，这是一个覆盖全部20种验证码类型的单一9B策略，采用监督微调后接强化学习。环境验证器直接提供强化学习奖励。监督微调达到70.5 Pass@1，强化学习进一步将其提升至71.7，同时也在两个外部基准上提升了性能。这些结果证明了大规模、细粒度的计算机使用监督对于训练交互式验证码智能体的价值。我们在该https URL发布了CaptchaArena和CaptchaAgent。
英文摘要
Interactive CAPTCHAs remain challenging for computer-use agents, while existing datasets face trade-offs among type coverage, interaction fidelity, and trajectory supervision. To address these gaps, we present CaptchaArena, the first large-scale, fine-grained training dataset for interactive CAPTCHA solving. It contains 50K puzzles across 20 CAPTCHA types and 5 interaction modes, with every solution verified through execution. CaptchaArena provides 50K screenshot-action trajectories, including 46K with step-by-step reasoning annotations. It also includes fine-grained pixel-mask annotations for irregular targets. Using CaptchaArena, we train CaptchaAgent, a single 9B policy for all 20 CAPTCHA types, with supervised fine-tuning followed by reinforcement learning. The environment verifier directly provides the RL reward. Supervised fine-tuning reaches 70.5 Pass@1, and reinforcement learning further improves it to 71.7, while also improving performance on two external benchmarks. These results demonstrate the value of large-scale, fine-grained computer-use supervision for training interactive CAPTCHA agents. We release CaptchaArena and CaptchaAgent at https://github.com/X0X0X00/CaptchaArena.
Comments30 pages, 4 figures, 16 tables. Code and data: https://github.com/X0X0X00/CaptchaArena
