cs.LG（2024-12-19）
📊 共 22 篇论文 | 🔗 1 篇有代码
🎯 兴趣领域导航
支柱二：RL算法与架构 (RL & Architecture) (11 🔗1)
支柱九：具身大模型 (Embodied Foundation Models) (7)
支柱八：物理动画 (Physics-based Animation) (4)
🔬 支柱二：RL算法与架构 (RL & Architecture) (11 篇)
| # | 题目 | 一句话要点 | 标签 | 🔗 | ⭐ |
|---|---|---|---|---|---|
| 1 | Cal-DPO: Calibrated Direct Preference Optimization for Language Model Alignment | 提出Cal-DPO，通过校准隐式奖励优化语言模型对齐，提升人类偏好对齐效果。 | DPO direct preference optimization large language model | ||
| 2 | AdaCred: Adaptive Causal Decision Transformers with Feature Crediting | AdaCred：基于特征可信度自适应因果决策Transformer，提升离线强化学习效率 | reinforcement learning offline RL offline reinforcement learning | ||
| 3 | GFormer: Accelerating Large Language Models with Optimized Transformers on Gaudi Processors | GFormer：利用优化Transformer加速Gaudi处理器上的大语言模型 | linear attention large language model | ||
| 4 | MARIA: a Multimodal Transformer Model for Incomplete Healthcare Data | 提出MARIA模型，解决医疗多模态数据缺失下的诊断与预测难题。 | predictive model multimodal | ||
| 5 | Hierarchical Subspaces of Policies for Continual Offline Reinforcement Learning | HiSPO：面向离线持续强化学习，利用分层策略子空间解决导航任务中的知识遗忘问题。 | reinforcement learning offline reinforcement learning | ||
| 6 | Entropy Regularized Task Representation Learning for Offline Meta-Reinforcement Learning | 提出熵正则化任务表征学习方法，提升离线元强化学习泛化能力 | reinforcement learning representation learning | ||
| 7 | Offline Safe Reinforcement Learning Using Trajectory Classification | 提出基于轨迹分类的离线安全强化学习方法，解决现有方法保守或违反约束问题 | reinforcement learning | ||
| 8 | MIETT: Multi-Instance Encrypted Traffic Transformer for Encrypted Traffic Classification | 提出MIETT：一种用于加密流量分类的多示例加密流量Transformer模型 | contrastive learning foundation model | ✅ | |
| 9 | Knowledge Distillation in RNN-Attention Models for Early Prediction of Student Performance | 提出RNN-Attention-KD框架，用于早期预测学生学业表现，助力教育干预。 | distillation | ||
| 10 | CLDG: Contrastive Learning on Dynamic Graphs | 提出CLDG框架，通过对比学习动态图中的时间平移不变性，提升无监督节点表示。 | contrastive learning | ||
| 11 | ST-ReP: Learning Predictive Representations Efficiently for Spatial-Temporal Forecasting | 提出ST-ReP模型，高效学习时空预测的表征，提升预测精度和可扩展性 | representation learning contrastive learning |
🔬 支柱九：具身大模型 (Embodied Foundation Models) (7 篇)
| # | 题目 | 一句话要点 | 标签 | 🔗 | ⭐ |
|---|---|---|---|---|---|
| 12 | Time Will Tell: Timing Side Channels via Output Token Count in Large Language Models | 利用LLM输出token数量的时间侧信道泄露推理输入敏感信息 | large language model | ||
| 13 | Bag of Tricks for Multimodal AutoML with Image, Text, and Tabular Data | 针对图像、文本和表格数据的多模态AutoML的经验总结与最佳实践 | multimodal | ||
| 14 | Large Language Models on Small Resource-Constrained Systems: Performance Characterization, Analysis and Trade-offs | 针对资源受限系统，论文评估并优化大语言模型在Jetson Orin上的性能表现。 | large language model | ||
| 15 | Data for Mathematical Copilots: Better Ways of Presenting Proofs for Machine Learning | 针对数学Copilot，提出更优的证明呈现方式，以提升机器学习效果。 | large language model | ||
| 16 | HPC-Coder-V2: Studying Code LLMs Across Low-Resource Parallel Languages | HPC-Coder-V2：研究低资源并行语言上的代码大语言模型，提升并行代码生成能力。 | large language model | ||
| 17 | Rethinking Uncertainty Estimation in Natural Language Generation | 提出G-NLL，通过单次贪婪解码实现高效可靠的自然语言生成不确定性估计 | large language model | ||
| 18 | GenAIOps for GenAI Model-Agility | 提出GenAIOps方法，应对生成式AI模型快速迭代带来的应用质量退化问题。 | foundation model |
🔬 支柱八：物理动画 (Physics-based Animation) (4 篇)
| # | 题目 | 一句话要点 | 标签 | 🔗 | ⭐ |
|---|---|---|---|---|---|
| 19 | LISA: Learning-Integrated Space Partitioning Framework for Traffic Accident Forecasting on Heterogeneous Spatiotemporal Data | 提出LISA框架，用于异构时空数据上的交通事故事故预测，实现自适应空间划分。 | spatiotemporal | ||
| 20 | A Generative Framework for Probabilistic, Spatiotemporally Coherent Downscaling of Climate Simulation | 提出基于扩散模型的生成框架，用于气候模拟的时空一致性降尺度 | spatiotemporal | ||
| 21 | GeoPro-Net: Learning Interpretable Spatiotemporal Prediction Models through Statistically-Guided Geo-Prototyping | GeoPro-Net：通过统计引导的地理原型学习可解释的时空预测模型 | spatiotemporal | ||
| 22 | A Universal Model for Human Mobility Prediction | 提出UniMob，一个通用的人类移动预测模型，可同时处理个体轨迹和群体流量预测。 | spatiotemporal |