Tag: Generalization
All the papers with the tag "Generalization".
Arbitrarily Applicable Same/Opposite Relational Responding with NARS
grok-3-latestScore: 0.48Published: at 18:03本文在非公理推理系统(NARS)中通过‘获取关系’机制,首次实现了任意适用的‘相同/相反’关系推导,结合对称性和传递性,为通用人工智能提供了接近人类符号推理的计算模型。
Rethinking Invariance in In-context Learning
grok-3-latestScore: 0.66Published: at 06:59本文提出 Invariant In-Context Learning (InvICL) 算法,通过设计不变性注意力掩码和两阶段编码策略,实现上下文学习对顺序的不变性,同时确保信息不泄露和上下文相互依赖,显著提升性能和泛化能力。
X-Reasoner: Towards Generalizable Reasoning Across Modalities and Domains
grok-3-latestScore: 0.48Published: at 21:08本文提出 X-REASONER,通过仅基于通用领域文本的两阶段后训练策略(SFT + RL),成功实现推理能力跨模态和跨领域泛化,并在多个通用和医学基准测试中超越现有 SOTA。
Patterns and Mechanisms of Contrastive Activation Engineering
grok-3-latestScore: 0.58Published: at 05:15本文系统分析了 Contrastive Activation Engineering (CAE) 的模式与机制,发现其在分布内有效但分布外泛化不足,且对模型困惑度有负面影响,为实际应用提供了重要参考。
Enhancing LLMs' Clinical Reasoning with Real-World Data from a Nationwide Sepsis Registry
grok-3-latestScore: 0.50Published: at 15:23本文通过真实世界脓毒症登记数据训练大型语言模型,显著提升其临床推理能力,并展示跨任务、跨疾病的泛化性,为通用临床推理模型的发展奠定基础。
On the generalization of language models from in-context learning and finetuning: a controlled study
grok-3-latestScore: 0.84Published: at 17:02本文通过控制实验揭示上下文学习在系统性泛化任务上优于微调,并提出通过上下文推理增强微调数据的方法,显著提升了微调的泛化能力。