Soohak让顶级模型只拿到30%

重点关注

• CollabVR: Collaborative Video Reasoning with Vision-Language and Video Generation Models score 10
  入选 HF Daily Papers；HF 热度: 46 upvotes (+4)；有代码实现；关键词(5): scaling, fine-tuning, reasoning, vision-language, open-source
• RigidFormer: Learning Rigid Dynamics using Transformers score 10
  机构: MIT；入选 HF Daily Papers；HF 热度: 8 upvotes (+2)；有代码实现
• LLaVA-UHD v4: What Makes Efficient Visual Encoding in MLLMs? score 9
  入选 HF Daily Papers；HF 热度: 11 upvotes (+3)；有代码实现；关键词(1): compression
• Soohak: A Mathematician-Curated Benchmark for Evaluating Research-level Math Capabilities of LLMs score 8
  入选 HF Daily Papers；HF 热度: 61 upvotes (+4)；关键词(1): reasoning
• MLS-Bench: A Holistic and Rigorous Assessment of AI Systems on Building Better AI score 8
  入选 HF Daily Papers；HF 热度: 6 upvotes (+2)；有代码实现；关键词(3): scaling, coding, reasoning
• AgentForesight: Online Auditing for Early Failure Prediction in Multi-Agent Systems score 7
  入选 HF Daily Papers；HF 热度: 4 upvotes (+1)；有代码实现；关键词(3): deployment, agentic, coding
• SlimQwen: Exploring the Pruning and Distillation in Large MoE Model Pre-training score 6
  入选 HF Daily Papers；HF 热度: 7 upvotes (+2)；关键词(6): compression, distillation, pruning, pre-training, pretraining
• FORTIS: Benchmarking Over-Privilege in Agent Skills score 5
  入选 HF Daily Papers；有代码实现

也值得关注

• S2FT: Parameter-Efficient Fine-Tuning in Sparse Spectrum Domain score 4
  关键词(1): fine-tuning；顶会接收: CVPR
• LightAVSeg: Lightweight Audio-Visual Segmentation score 4
  关键词(2): lightweight, deployment；顶会接收: ICML
• FRACTAL: SSM with Fractional Recurrent Architecture for Computational Temporal Analysis of Long Sequences score 4
  关键词(1): state space；顶会接收: ICML
• ReST-KV: Robust KV Cache Eviction with Layer-wise Output Reconstruction and Spatial-Temporal Smoothing score 4
  关键词(1): latency；顶会接收: ICLR
• Illusion-Aware Visual Preprocessing and Anti-Illusion Prompting for Classic Illusion Understanding in Vision-Language Models score 4
  关键词(3): edge, finetuning, vision-language；顶会接收: CVPR
• OTora: A Unified Red Teaming Framework for Reasoning-Level Denial-of-Service in LLM Agents score 4
  关键词(2): latency, reasoning；顶会接收: ICML
• BoostAPR: Boosting Automated Program Repair via Execution-Grounded Reinforcement Learning with Dual Reward Models score 4
  关键词(4): fine-tuning, PPO, reasoning, open-source；顶会接收: ICML
• Structure-Centric Graph Foundation Model via Geometric Bases score 3
  顶会接收: ICML
• SynerMedGen: Synergizing Medical Multimodal Understanding with Generation via Task Alignment score 3
  顶会接收: ICML
• Unlearners Can Lie: Evaluating and Improving Honesty in LLM Unlearning score 3
  顶会接收: ACL
• Geometrically Constrained Stenosis Editing in Coronary Angiography via Entropic Optimal Transport score 3
  顶会接收: ICML