Agent红队11种失败模式，step级路由降本700倍

重点关注

• A Very Big Video Reasoning Suite score 9
  入选 HF Daily Papers；HF 热度: 404 upvotes (+4)；关键词(4): scaling, reasoning, benchmark, evaluation
• SkillOrchestra: Learning to Route Agents via Skill Transfer score 9
  入选 HF Daily Papers；HF 热度: 47 upvotes (+4)；关键词(5): efficient, deployment, agent, agents, cost
• Agents of Chaos score 9
  入选 HF Daily Papers；HF 热度: 24 upvotes (+4)；关键词(5): deployment, attention, agent, agents, tool use
• ManCAR: Manifold-Constrained Latent Reasoning with Adaptive Test-Time Computation for Sequential Recommendation score 9
  入选 HF Daily Papers；HF 热度: 23 upvotes (+4)；关键词(2): reasoning, recommendation
• Mobile-O: Unified Multimodal Understanding and Generation on Mobile Device score 9
  入选 HF Daily Papers；HF 热度: 21 upvotes (+4)；关键词(10): efficient, efficiency, deployment, real-time, edge
• tttLRM: Test-Time Training for Long Context and Autoregressive 3D Reconstruction score 9
  入选 HF Daily Papers；HF 热度: 4 upvotes (+1)；关键词(3): scaling, fast, pretraining；顶会接收: CVPR
• DSDR: Dual-Scale Diversity Regularization for Exploration in LLM Reasoning score 7
  入选 HF Daily Papers；HF 热度: 11 upvotes (+3)；关键词(1): reasoning
• QuantVLA: Scale-Calibrated Post-Training Quantization for Vision-Language-Action Models score 7
  入选 HF Daily Papers；HF 热度: 8 upvotes (+2)；关键词(13): scaling, lightweight, quantization, deployment, inference
• TAPE: Tool-Guided Adaptive Planning and Constrained Execution in Language Model Agents score 7
  入选 HF Daily Papers；HF 热度: 5 upvotes (+2)；关键词(3): agent, agents, planning
• Implicit Intelligence -- Evaluating Agents on What Users Don't Say score 6
  入选 HF Daily Papers；HF 热度: 3 upvotes (+1)；关键词(5): agent, agents, agentic, reasoning, evaluation

也值得关注

• Personalized Prediction of Perceived Message Effectiveness Using Large Language Model Based Digital Twins score 4
  机构: University of Washington；关键词(1): fine-tuning
• Model Merging in the Essential Subspace score 4
  关键词(1): scaling；顶会接收: CVPR
• Relational Feature Caching for Accelerating Diffusion Transformers score 4
  关键词(1): diffusion；顶会接收: ICLR
• Grokking Finite-Dimensional Algebra score 3
  机构: Mila
• What Matters for Simulation to Online Reinforcement Learning on Real Robots score 3
  机构: DeepMind
• Discover, Segment, and Select: A Progressive Mechanism for Zero-shot Camouflaged Object Segmentation score 3
  顶会接收: CVPR
• The Invisible Gorilla Effect in Out-of-distribution Detection score 3
  顶会接收: CVPR
• Conformal Risk Control for Non-Monotonic Losses score 3
  机构: Berkeley
• Hiding in Plain Text: Detecting Concealed Jailbreaks via Activation Disentanglement score 2
  关键词(3): inference, safety, jailbreak
• Hilbert-Augmented Reinforcement Learning for Scalable Multi-Robot Coverage and Exploration score 2
  关键词(3): efficiency, PPO, robotics
• LEVDA: Latent Ensemble Variational Data Assimilation via Differentiable Dynamics score 2
  关键词(2): efficient, efficiency
• Federated Causal Representation Learning in State-Space Systems for Decentralized Counterfactual Reasoning score 2
  关键词(2): inference, reasoning
• IR$^3$: Contrastive Inverse Reinforcement Learning for Interpretable Detection and Mitigation of Reward Hacking score 2
  关键词(3): distillation, RLHF, alignment
• RAmmStein: Regime Adaptation in Mean-reverting Markets with Stein Thresholds -- Optimal Impulse Control in Concentrated AMMs score 2
  关键词(3): efficiency, state space, agent
• Prefer-DAS: Learning from Local Preferences and Sparse Prompts for Domain Adaptive Segmentation of Electron Microscopy score 2
  关键词(4): efficient, inference, alignment, preference