Trillion-Parameter MoE Post-Training Runs on Huawei Ascend

Featured

• DocOps: A Verifiable Benchmark for Autonomous Agents in Complex Document Operations score 8
  入选 HF Daily Papers; HF 热度: 7 upvotes (+2); 有代码实现; 关键词(2): agentic, open-source
• SLAI T-Rex: Full-Parameter Post-training of the DeepSeek-V4 Family on Ascend SuperPOD score 10
  入选 HF Daily Papers; HF 热度: 55 upvotes (+4); 有代码实现; 关键词(4): post-training, MoE, reasoning, open-source
• Self Gradient Forcing: Native Long Video Extrapolation score 9
  入选 HF Daily Papers; HF 热度: 30 upvotes (+4); 有代码实现
• Beyond Relevance-Centric Retrieval: Rubric-Oriented Document Set Selection and Ranking score 9
  入选 HF Daily Papers; HF 热度: 26 upvotes (+4); 有代码实现
• Scaling Laws for Hypernetwork-Based Knowledge Injection in Large Language Models score 9
  入选 HF Daily Papers; HF 热度: 13 upvotes (+3); 有代码实现; 关键词(4): scaling, fine-tuning, finetuning, reasoning
• SLPO: Scaling Latent Reasoning via a Surrogate Policy score 7
  入选 HF Daily Papers; HF 热度: 4 upvotes (+1); 有代码实现; 关键词(2): scaling, reasoning
• Trace: A Taxonomy-Guided Environment for Multidomain Visual Reasoning score 8
  入选 HF Daily Papers; HF 热度: 5 upvotes (+2); 有代码实现; 关键词(2): reasoning, vision-language
• G-MAD: A Game-Based Data Generation Framework for Multi-View RGB-T Aerial Object Detection score 7
  入选 HF Daily Papers; HF 热度: 4 upvotes (+1); 有代码实现; 关键词(1): open-source
• Reading and Steering Representations of Materials-Science Mechanisms in an Open-Weight Language Model score 7
  入选 HF Daily Papers; HF 热度: 5 upvotes (+2); 有代码实现

Also Worth Noting

• ATSplat: Compact Feed-forward 3D Gaussian Splatting with Adaptive Token Expansion score 5
  入选 HF Daily Papers; HF 热度: 7 upvotes (+2)
• Train the Model, Not the Reader: Decodability Supervision for Verifiable Activation Explanations score 4
  入选 HF Daily Papers; HF 热度: 4 upvotes (+1)
• ENTRAP-VL: A Taxonomic Probe for Dual Contextual Entrainment in Vision-Language Models score 5
  入选 HF Daily Papers; HF 热度: 2 upvotes (+1); 关键词(1): vision-language