一篇探讨对比学习框架下弱到强泛化的新论文，理论分析和实验验证结合，为AI大模型泛化研究提供新视角。
arXiv:2510.07884v2 Announce Type: replace-cross Abstract: Weak-to-strong generalization provides a promising paradigm for scaling large language model…
LLM智能体安全新方案：通过自演进对比记忆机制，让防御系统在动态攻击中持续强化，兼具可塑性与鲁棒性。
arXiv:2606.05743v1 Announce Type: cross Abstract: Despite advances in safety alignment, large language models remain vulnerable to continuously evolvi…
用对比学习训练神经网络来模仿图着色算法，52页论文揭秘神经推理新突破。
arXiv:2606.03923v1 Announce Type: new Abstract: Graph coloring seeks to assigns colors to a graph's nodes so that adjacent nodes receive different col…
热感知融入CLIP，多模态视觉语言预训练新突破，拓展模型在低光照等场景的感知能力。
arXiv:2606.00673v1 Announce Type: new Abstract: Thermal imaging offers a powerful alternative to visible-spectrum vision under challenging conditions …
提出对比表示正则化方法，有效提升VLA模型在机器人操作中对动作信号的敏感性，弥补现有表示缺陷。
arXiv:2510.01711v3 Announce Type: replace-cross Abstract: Vision-Language-Action (VLA) models have shown strong capabilities in robot manipulation by …
LLM Agent自主内化技能的新方法：对比信用分配让Agent学会评估行动贡献，提升复杂任务表现
arXiv:2605.27899v1 Announce Type: new Abstract: Structured skill prompts improve exploration in long-horizon agentic reinforcement learning (RL). Skil…
仅用LLM和对比估计提示就能预测点互信息，PromptNCE方法简洁高效，值得关注。
arXiv:2605.21776v1 Announce Type: new Abstract: Estimating mutual information from text usually requires training a task-specific critic, which limits…
提出TiTok方法，通过对比学习转移Token级知识，让LoRA参数可在不同骨干模型间移植。
arXiv:2510.04682v3 Announce Type: replace-cross Abstract: Large Language Models (LLMs) are widely applied in real world scenarios, yet fine-tuning the…
结合视觉核心引导与对比学习，攻克多模态中风预后预测的平衡难题。
arXiv:2605.14710v1 Announce Type: cross Abstract: Deep learning and multi-modal fusion have demonstrated transformative potential in medical diagnosis…
新方法AudioMosaic融合对比学习与掩码表示，攻克音频自监督的对比方法难题。
arXiv:2605.14231v1 Announce Type: cross Abstract: Audio self-supervised learning (SSL) aims to learn general-purpose representations from large-scale …