Open-Source Deep Research Beats GPT-5, Embedding Scaling Outshines Experts

Featured

• Grounding and Enhancing Informativeness and Utility in Dataset Distillation score 11
  入选 HF Daily Papers; HF 热度: 15 upvotes (+3); 关键词(3): efficiency, distillation, synthetic data; 顶会接收: ICLR
• Hybrid Linear Attention Done Right: Efficient Distillation and Effective Architectures for Extremely Long Contexts score 8
  入选 HF Daily Papers; HF 热度: 10 upvotes (+3); 关键词(9): efficient, efficiency, distillation, inference, throughput
• Vision-DeepResearch: Incentivizing DeepResearch Capability in Multimodal Large Language Models score 9
  入选 HF Daily Papers; HF 热度: 147 upvotes (+4); 关键词(3): reasoning, multimodal, search
• Scaling Embeddings Outperforms Scaling Experts in Language Models score 9
  入选 HF Daily Papers; HF 热度: 97 upvotes (+4); 关键词(5): scaling, inference, MoE, agentic, coding
• CAR-bench: Evaluating the Consistency and Limit-Awareness of LLM Agents under Real-World Uncertainty score 9
  入选 HF Daily Papers; HF 热度: 69 upvotes (+4); 关键词(4): agents, tool use, reasoning, benchmark
• DynamicVLA: A Vision-Language-Action Model for Dynamic Object Manipulation score 9
  入选 HF Daily Papers; HF 热度: 67 upvotes (+4); 关键词(8): efficient, fast, inference, latency, reasoning
• MMFineReason: Closing the Multimodal Reasoning Gap via Open Data-Centric Methods score 9
  入选 HF Daily Papers; HF 热度: 57 upvotes (+4); 关键词(5): efficiency, fine-tune, reasoning, multimodal, open-source
• ASTRA: Automated Synthesis of agentic Trajectories and Reinforcement Arenas score 9
  入选 HF Daily Papers; HF 热度: 56 upvotes (+4); 关键词(5): efficiency, fine-tuning, agents, agentic, reasoning
• OCRVerse: Towards Holistic OCR in End-to-End Vision-Language Models score 9
  入选 HF Daily Papers; HF 热度: 49 upvotes (+4); 关键词(3): multimodal, vision-language, open-source
• ConceptMoE: Adaptive Token-to-Concept Compression for Implicit Compute Allocation score 9
  入选 HF Daily Papers; HF 热度: 42 upvotes (+4); 关键词(9): efficiency, compression, pretraining, attention, MoE

Also Worth Noting

• Mechanistic Data Attribution: Tracing the Training Origins of Interpretable LLM Units score 4
  入选 HF Daily Papers; HF 热度: 4 upvotes (+1)
• Distributionally Robust Classification for Multi-source Unsupervised Domain Adaptation score 4
  关键词(1): efficient; 顶会接收: ICLR
• Do Pathology Foundation Models Encode Disease Progression? A Pseudotime Analysis of Visual Representations score 4
  机构: Harvard; 关键词(1): diffusion
• Bi-Anchor Interpolation Solver for Accelerating Generative Modeling score 4
  关键词(2): lightweight, latency; GitHub 热门仓库
• CORDS: Continuous Representations of Discrete Structures score 4
  关键词(1): inference; 顶会接收: ICLR
• A Federated Generalized Expectation-Maximization Algorithm for Mixture Models with an Unknown Number of Components score 3
  顶会接收: ICLR
• Optimal Transport-Induced Samples against Out-of-Distribution Overconfidence score 3
  顶会接收: ICLR
• Beyond a Single Reference: Training and Evaluation with Paraphrases in Sign Language Translation score 2
  关键词(2): alignment, evaluation
• Large Language Models Naively Recover Ethnicity from Individual Records score 2
  关键词(6): deployment, inference, transformer, reasoning, open-source
• Optimization and Mobile Deployment for Anthropocene Neural Style Transfer score 2
  关键词(5): scaling, deployment, inference, latency, real-time
• BrainStack: Neuro-MoE with Functionally Guided Expert Routing for EEG-Based Language Decoding score 2
  关键词(4): distillation, transformer, MoE, benchmark