1
Recursive Synthesis for Long-Horizon Terminal Tasks
面向长时程终端任务，递归合成方法带来全新解决思路，值得关注。
arXiv:2608.05466v1 Announce Type: new Abstract: High-quality long-horizon training data for terminal agents is expensive to produce, often costing hun…
2
TRAM: Enhancing Multimodal Reasoning with Trajectory-Derived Auxiliary Memory
用轨迹生成辅助记忆，多模态推理能力再进一步，值得关注的新方法。
arXiv:2608.01922v1 Announce Type: new Abstract: Multimodal Large Reasoning Models (MLRMs) have achieved strong performance on tasks requiring visual u…
3
Enhancing Rubric-based RL via Self-Distillation
用自蒸馏技术增强基于评分标准的强化学习，提升训练效果和泛化能力。
arXiv:2607.18082v1 Announce Type: cross Abstract: Rubric-based RL has recently shown promise in improving LLMs on open-ended tasks. A widely recognize…
4
Imagine Before You Predict: Interleaved Latent Visual Reasoning for Video Event Prediction
先想象再预测：论文提出交错潜在视觉推理新方法，提升视频事件预测准确性与可解释性。
arXiv:2606.05769v1 Announce Type: new Abstract: Video event prediction (VEP) requires models to infer unobserved future states from partial video evid…