cs.LG(2025-12-26)

📊 共 8 篇论文 | 🔗 3 篇有代码

🎯 兴趣领域导航

支柱二:RL算法与架构 (RL & Architecture) (4 🔗1) 支柱九:具身大模型 (Embodied Foundation Models) (3 🔗1) 支柱四:生成式动作 (Generative Motion) (1 🔗1)

🔬 支柱二:RL算法与架构 (RL & Architecture) (4 篇)

#题目一句话要点标签🔗
1 MMCTOP: A Multimodal Textualization and Mixture-of-Experts Framework for Clinical Trial Outcome Prediction 提出MMCTOP框架以解决多模态临床试验结果预测问题 representation learning multimodal
2 Exploring the Heterogeneity of Tabular Data: A Diversity-aware Data Generator via LLMs 提出DATE框架,利用LLM生成多样性表格数据,提升小样本学习性能。 DPO direct preference optimization large language model
3 Semiparametric Preference Optimization: Your Language Model is Secretly a Single-Index Model 提出半参数偏好优化方法,解决语言模型对齐中链接函数未知的问题 policy learning large language model
4 A Comedy of Estimators: On KL Regularization in RL Training of LLMs 研究KL散度估计器对LLM的RL训练影响,提升模型在分布内外的泛化性能。 reinforcement learning large language model

🔬 支柱九:具身大模型 (Embodied Foundation Models) (3 篇)

#题目一句话要点标签🔗
5 Explainable Multimodal Regression via Information Decomposition 提出基于信息分解的可解释多模态回归框架,提升预测精度与可解释性。 multimodal
6 Unifying Learning Dynamics and Generalization in Transformers Scaling Law 提出统一学习动态与变压器缩放法则以提升模型泛化能力 large language model
7 Prefill vs. Decode Bottlenecks: SRAM-Frequency Tradeoffs and the Memory-Bandwidth Ceiling 研究SRAM频率权衡与内存带宽瓶颈,优化LLM推理能效 large language model

🔬 支柱四:生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗
8 GQ-VAE: A gated quantized VAE for learning variable length tokens 提出门控量化VAE(GQ-VAE),用于学习变长token,作为现有tokenizer的即插即用替代方案。 VQ-VAE

⬅️ 返回 cs.LG 首页 · 🏠 返回主页