论文来源 | 控制协议达100%，循环层省18%计算

重点关注

StudentSim: Training LLM-based Student Simulators score 9
入选 HF Daily Papers；HF 热度: 478 upvotes (+4)；有代码实现
Evaluating Multimodal LLMs as Generalist Vision-Language-Action Agents for Drone Control: Commanding, Approaching, Tracking and Searching score 10
入选 HF Daily Papers；HF 热度: 24 upvotes (+4)；有代码实现；关键词(4): edge, fine-tuning, vision-language, embodied
H3-World: Turning Language Understanding into World Control score 10
入选 HF Daily Papers；HF 热度: 47 upvotes (+4)；有代码实现；关键词(2): lightweight, pretraining
EM^2Mem: Event-Centric Multimodal Memory for Large Language Models score 7
入选 HF Daily Papers；HF 热度: 12 upvotes (+3)；关键词(1): latency
SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers score 8
入选 HF Daily Papers；HF 热度: 86 upvotes (+4)；关键词(2): scaling, MoE
Control-Data Flow Separation: Stable Prompt Optimization in Multi-Agent LLMs score 8
入选 HF Daily Papers；HF 热度: 8 upvotes (+2)；有代码实现；关键词(1): reasoning
Uncovering Understanding-Generation Synergy in Native Unified Multimodal Models: From Representation, Task to System score 7
入选 HF Daily Papers；HF 热度: 23 upvotes (+4)
The Safeguard Worked. Is the LLM System Safer? score 5
入选 HF Daily Papers；HF 热度: 4 upvotes (+1)；关键词(1): deployment
Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement score 7
入选 HF Daily Papers；HF 热度: 13 upvotes (+3)；关键词(2): deployment, coding
ReFlowSET: Representation-Aligned Latent Flow Matching for SAR-to-EO Image Translation score 7
入选 HF Daily Papers；HF 热度: 5 upvotes (+2)；有代码实现

也值得关注

Runtime-Independent Persistent Agents: Preserving Identity, Memory, and Code Across Models, Harnesses, and Servers score 4
机构: University of Washington；关键词(1): deployment
Aligned but Flattened: Analyzing the Trade-off between Cultural Alignment and Diversity in LLMs score 4
关键词(2): fine-tuning, post-training；顶会接收: EMNLP
DramaChain Bench: An End-to-End Benchmark for Short-Drama Generation score 9
机构: Tencent；入选 HF Daily Papers；HF 热度: 5 upvotes (+2)；关键词(2): production, agentic