Code Agents Can't Cross Repo Boundaries, Under 45% Success

Featured

• CFG-Ctrl: Control-Based Classifier-Free Diffusion Guidance score 11
  入选 HF Daily Papers; HF 热度: 6 upvotes (+2); 有代码实现; 关键词(1): text-to-image; 顶会接收: CVPR
• Utonia: Toward One Encoder for All Point Clouds score 10
  入选 HF Daily Papers; HF 热度: 133 upvotes (+4); 有代码实现; 关键词(4): reasoning, vision-language, robotics, embodied
• UniG2U-Bench: Do Unified Models Advance Multimodal Understanding? score 10
  入选 HF Daily Papers; HF 热度: 78 upvotes (+4); 有代码实现; 关键词(3): pretraining, reasoning, vision-language
• BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? score 10
  入选 HF Daily Papers; HF 热度: 50 upvotes (+4); 有代码实现; 关键词(2): coding, reasoning
• MemSifter: Offloading LLM Memory Retrieval via Outcome-Driven Proxy Reasoning score 10
  入选 HF Daily Papers; HF 热度: 20 upvotes (+4); 有代码实现; 关键词(1): reasoning
• NOVA: Sparse Control, Dense Synthesis for Pair-Free Video Editing score 10
  入选 HF Daily Papers; HF 热度: 7 upvotes (+2); 有代码实现; 顶会接收: CVPR
• PRISM: Pushing the Frontier of Deep Think via Process Reward Model-Guided Inference score 9
  入选 HF Daily Papers; HF 热度: 18 upvotes (+3); 有代码实现; 关键词(1): reasoning
• Next Embedding Prediction Makes World Models Stronger score 9
  入选 HF Daily Papers; HF 热度: 16 upvotes (+3); 有代码实现; 关键词(1): reasoning
• Heterogeneous Agent Collaborative Reinforcement Learning score 8
  入选 HF Daily Papers; HF 热度: 92 upvotes (+4); 关键词(3): distillation, deployment, reasoning
• Beyond Language Modeling: An Exploration of Multimodal Pretraining score 8
  入选 HF Daily Papers; HF 热度: 65 upvotes (+4); 关键词(3): scaling, pretraining, MoE

Also Worth Noting

• Through the Lens of Contrast: Self-Improving Visual Reasoning in VLMs score 4
  关键词(2): finetuning, reasoning; 顶会接收: ICLR
• Agentified Assessment of Logical Reasoning Agents score 4
  关键词(1): reasoning; 顶会接收: ICLR
• SIGMark: Scalable In-Generation Watermark with Blind Extraction for Video Diffusion score 4
  关键词(1): coding; 顶会接收: ICLR
• LLandMark: A Multi-Agent Framework for Landmark-Aware Multimodal Interactive Video Retrieval score 4
  关键词(1): reasoning; 顶会接收: AAAI
• Learning to Generate and Extract: A Multi-Agent Collaboration Framework For Zero-shot Document-level Event Arguments Extraction score 4
  关键词(1): synthetic data; 顶会接收: AAAI
• MoECLIP: Patch-Specialized Experts for Zero-shot Anomaly Detection score 4
  关键词(1): MoE; 顶会接收: CVPR
• Inherited Goal Drift: Contextual Pressure Can Undermine Agentic Goals score 4
  关键词(3): deployment, post-training, agentic; 顶会接收: ICLR
• When Shallow Wins: Silent Failures and the Depth-Accuracy Paradox in Latent Reasoning score 4
  关键词(2): scaling, reasoning; 顶会接收: ICLR
• The Controllability Trap: A Governance Framework for Military AI Agents score 4
  关键词(3): real-time, agentic, tool use; 顶会接收: ICLR
• Test-Time Meta-Adaptation with Self-Synthesis score 4
  关键词(2): reasoning, synthetic data; 顶会接收: ICLR
• Online Learnability of Chain-of-Thought Verifiers: Soundness and Completeness Trade-offs score 4
  机构: Carnegie Mellon; 关键词(1): reasoning
• SENTINEL: Stagewise Integrity Verification for Pipeline Parallel Decentralized Training score 4
  机构: Amazon; 关键词(1): lightweight
• Improving Diffusion Planners by Self-Supervised Action Gating with Energies score 3
  机构: Carnegie Mellon
• FiDeSR: High-Fidelity and Detail-Preserving One-Step Diffusion Super-Resolution score 3
  顶会接收: CVPR
• Rethinking Code Similarity for Automated Algorithm Design with LLMs score 3
  顶会接收: ICLR