cs.LG（2026-09-22）

📊 共 21 篇论文 | 🔗 2 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (11 🔗2) 支柱二：RL算法与架构 (RL & Architecture) (9) 支柱一：机器人控制 (Robot Control) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (11 篇)

#题目一句话要点标签🔗⭐
1 Interweaving Marginals into Multivariate Sample Paths: Training-Free Dependence Construction for Probabilistic Time Series Foundation Models 提出无训练依赖构建方法以增强概率时间序列模型的多变量预测能力 foundation model
2 Evaluating Accuracy and Probabilistic Reliability of Zero-Shot Time Series Foundation Models 提出基于零-shot时间序列模型的准确性与可靠性评估方法 foundation model
3 EquivSVA: A Formally Verified Dataset of Behavioral Assertions Across Equivalent RTL Implementations 提出EquivSVA以验证行为断言在等效RTL实现中的一致性 large language model ✅
4 Greedy Decoding Is Not Precision-Invariant: Cross-Precision Output Divergence in LLM Inference 提出精确度不变性分析以解决大语言模型推理中的输出差异问题 large language model
5 CompKV: Compensation-Aware KV Selection for Long-Context LLM Inference 提出CompKV以解决长上下文LLM推理中的KV缓存流量问题 large language model
6 Learning to Fluctuate: Statistical Foundations for Causal Tabular Pretraining 提出波动监督预训练以解决因果表格建模问题 foundation model
7 You Only Need 2/3 of the Chosen Experts: An Empirical Study of Dynamic Expert Pruning in Fine-Grained MoE LLMs 提出动态专家剪枝方法以提升细粒度MoE LLMs的推理效率 multimodal
8 Latest Exact Match Attention 提出最新精确匹配注意力机制以提升Transformer性能 chain-of-thought ✅
9 From Experts to Sub-experts: Fine-grained Parameter-Efficient Fine-Tuning for MoE LLMs 提出NSFT以实现MoE LLMs的细粒度参数高效微调 large language model
10 Accelerating the Mitigation of LLM Inference Nondeterminism Across GPU Architectures 提出固定配置的GEMM内核以解决GPU架构间LLM推理非确定性问题 large language model
11 Hill Sampling for Test-Time Scaling: A Simple and Better Alternative to Repeated Sampling, Evolution, and Training 提出Hill Sampling以优化测试时计算分配策略 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (9 篇)

#题目一句话要点标签🔗⭐
12 A JEPA Recipe for Tabular Foundation Models 提出JEPA架构以提升表格基础模型的预测能力 world model world models JEPA
13 PACT: From Credit Assignment to Critic Alignment 提出PACT以解决强化学习中的信用分配问题 reinforcement learning PPO distillation
14 What Should a Self-Teacher See? Privileged Context Design for On-Policy Self-Distillation 提出特权上下文设计以优化自我蒸馏效果 teacher-student distillation privileged information
15 MAGIC: Mixed-Granularity Agent Graphs via Incremental Construction with Dense-Reward Reinforcement Learning 提出MAGIC框架以解决多代理系统中的协作拓扑问题 reinforcement learning reward shaping
16 HYDRA: Proactive Android Malware Drift Adaptation via Hierarchical Graph Contrastive Learning 提出HYDRA框架以解决Android恶意软件概念漂移问题 contrastive learning
17 Gaussian Flow-Matching Schedules: Implications for Sampling and Training 提出高斯流匹配调度以优化采样与训练过程 flow matching
18 Fully Byzantine-Resilient Multi-Agent Reinforcement Learning 提出FRAC-MARL以解决分布式拜占庭鲁棒性问题 reinforcement learning
19 Graph Domain Adaptation Does Not End with Representation Learning 提出EviGDA框架以解决图域适应中的预测证据不足问题 representation learning
20 Optimal Sequential Annotations for Off-Policy Evaluation 提出优化序列标注方法以提升离线强化学习评估效果 reinforcement learning offline reinforcement learning

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
21 Minimal Recurrent Behavioral Memory for Imitation under Partial Observability 提出最小递归行为记忆以解决部分可观测下模仿问题 manipulation

⬅️ 返回 cs.LG 首页 · 🏠 返回主页