PAPER EVIDENCE RANKING
基于PaperMiner证据链动态评分。缺失证据保持未知，不等同于零分；不同任务论文不宜只凭总分直接比较。
| 排名 | 学校 | 得分 | 论文数 | 覆盖率 |
|---|---|---|---|---|
| 1 | Tsinghua University FuzzySQL: Uncovering Hidden Vulnerabilities in DBMS Special Features with LLM-Driven Fuzzing · PCIE Pose Solution for EgoExo4D Pose and Proficiency Estimation Challenge · Relit-LiVE: Relight Video by Jointly Learning Environment Video | 55.69 | 321 | 94% |
| 2 | 上海交通大学 AnchorSync: Global Consistency Optimization for Long Video Editing · GraphDLG: Exploring Deep Leakage from Gradients in Federated Graph Learning · MoGen: A Unified Collaborative Framework for Controllable Multi-Object Image Generation | 55.69 | 274 | 94% |
| 3 | Zhejiang University NeurIDA: Dynamic Modeling for Effective In-Database Analytics · Generative Video Matting · FT-RAG: A Fine-grained Retrieval-Augmented Generation Framework for Complex Table Reasoning | 55.69 | 246 | 95% |
| 4 | Peking University LoBoFit: Flexible Garment Refitting via Local Bone Mapping Blending · Crystal structure prediction with nuclear quantum and finite-temperature effects via deep free energy learning · FilDeep: Learning Large Deformations of Elastic-Plastic Solids with Multi-Fidelity Data | 55.69 | 227 | 92% |
| 5 | University of California DeltaDorsal: Enhancing Hand Pose Estimation with Dorsal Features in Egocentric Views · Predicting food taste with bound-driven optimization · FAIRCAUSESYN: TOWARDS CAUSALLY FAIR LLM-AUGMENTED SYNTHETIC DATA GENERATION | 55.69 | 223 | 92% |
| 6 | University of Science and Technology of China LoBoFit: Flexible Garment Refitting via Local Bone Mapping Blending · DeepGen 1.0: A Lightweight Unified Multimodal Model for Advancing Image Generation and Editing · InstructTime $^{++}$ : Time Series Classification with Multimodal Language Modeling via Implicit Feature Enhancement | 55.69 | 183 | 92% |
| 7 | Nanyang Technological University Temporal Restoration and Spatial Rewiring for Source-Free Multivariate Time Series Domain Adaptation · The Last Human-Written Paper: Agent-Native Research Artifacts · Imagine360: Immersive 360 Video Generation from Perspective Anchor | 55.69 | 165 | 91% |
| 8 | The Hong Kong University of Science and Technology Relit-LiVE: Relight Video by Jointly Learning Environment Video · Rewarding Graph Reasoning Process makes LLMs more Generalized Reasoners · Agentic Enterprise: AI-Centric User to User-Centric AI | 55.69 | 162 | 94% |
| 9 | Fudan University DeepGen 1.0: A Lightweight Unified Multimodal Model for Advancing Image Generation and Editing · From Understanding to Action: Feedback-Grounded Policy Discovery for Generative Recommendation · TensoFlow: Tensorial Flow-based Sampler for Inverse Rendering | 55.69 | 154 | 91% |
| 10 | National University of Singapore Factorized Learning for Temporally Grounded Video-Language Models · NeurIDA: Dynamic Modeling for Effective In-Database Analytics · High-Resolution Be Aware! Improving the Self-Supervised Real-World Super-Resolution | 55.69 | 147 | 94% |
| 11 | The Chinese University of Hong Kong Approaching Low-Cost Cardiac Intelligence with Semi-Supervised Knowledge Distillation · Relit-LiVE: Relight Video by Jointly Learning Environment Video · Unifying Physically-Informed Weather Priors in A Single Model for Image Restoration Across Multiple Adverse Weather Conditions | 55.69 | 145 | 92% |
| 12 | Stanford University Latent Chain-of-Thought World Modeling for End-to-End Autonomous Driving · What LLMs Think When You Don’t Tell Them What to Think About? · The Last Human-Written Paper: Agent-Native Research Artifacts | 55.69 | 142 | 92% |
| 13 | University of Chinese Academy of Sciences PCIE Pose Solution for EgoExo4D Pose and Proficiency Estimation Challenge · Relit-LiVE: Relight Video by Jointly Learning Environment Video · MGPC: Multimodal Network for Generalizable Point Cloud Completion With Modality Dropout and Progressive Decoding | 55.69 | 139 | 92% |
| 14 | Massachusetts Institute of Technology BRAIN: Bias-Mitigation Continual Learning Approach to Vision-Brain Understanding · The Last Human-Written Paper: Agent-Native Research Artifacts · Dark matter searches with a 13 meV threshold superconducting sensor array | 55.69 | 138 | 92% |
| 15 | Harbin Institute of Technology ObjFiller-3D: Consistent Multi-view 3D Inpainting via Video Diffusion Models · HoloTime: Taming Video Diffusion Models for Panoramic 4D Scene Generation · SpatialGrammar: A Domain-Specific Language for LLM-Based 3D Indoor Scene Generation | 55.69 | 119 | 90% |
| 16 | Carnegie Mellon University The Last Human-Written Paper: Agent-Native Research Artifacts · Combining e-values using demi-supermartingales · COMPARING AI AGENTS TO CYBERSECURITY PROFESSIONALS IN REAL-WORLD PENETRATION TESTING | 55.69 | 105 | 91% |
| 17 | Nanjing University DeepGen 1.0: A Lightweight Unified Multimodal Model for Advancing Image Generation and Editing · Relit-LiVE: Relight Video by Jointly Learning Environment Video · Towards Perfection: Building Inter-component Mutual Correction for Retinex-based Low-light Image Enhancement | 55.69 | 89 | 92% |
| 18 | Imperial College SPEAR: A Simulator for Photorealistic Embodied AI Research · SPAC: Automating FPGA-based Network Switches with Protocol Adaptive Customization · Semantic Sensing: A Task-Oriented Paradigm | 55.69 | 85 | 90% |
| 19 | Beihang University GraphDLG: Exploring Deep Leakage from Gradients in Federated Graph Learning · Seeing through Light and Darkness: Sensor-Physics Grounded Deblurring HDR NeRF from Single-Exposure Images and Events · FS-Diff: Semantic Guidance and Clarity-Aware Simultaneous Multimodal Image Fusion and Super-Resolution | 55.69 | 84 | 90% |
| 20 | The University of Hong Kong When to Retrieve During Reasoning: Adaptive Retrieval for Large Reasoning Models · Rein3D: Reinforced 3D Indoor Scene Generation with Panoramic Video Diffusion Models · CamDirector: Towards Long-Term Coherent Video Trajectory Editing | 55.69 | 80 | 90% |
| 21 | Sun Yat-sen University ObjFiller-3D: Consistent Multi-view 3D Inpainting via Video Diffusion Models · CFSR: Geometry-Conditioned Shadow Removal via Physical Disentanglement · DEPTH-GUIDED METRIC-AWARE TEMPORAL CONSISTENCY FOR MONOCULAR VIDEO HUMAN MESH RECOVERY | 55.69 | 76 | 90% |
| 22 | The Hong Kong Polytechnic University High-resolution Photo Enhancement in Real-time: A Laplacian Pyramid Network · CredibleDFGO: Differentiable Factor Graph Optimization with Credibility Supervision · CheatAgent: Attacking LLM-Empowered Recommender Systems via LLM Agent | 55.69 | 66 | 92% |
| 23 | Huazhong University of Science and Technology High-resolution Photo Enhancement in Real-time: A Laplacian Pyramid Network · Relit-LiVE: Relight Video by Jointly Learning Environment Video · FaGeL: Fabric LLMs Agent empowered Embodied Intelligence Evolution with Autonomous Human-Machine Collaboration | 55.69 | 65 | 91% |
| 24 | Beijing Institute of Technology MM-DETR: An Efficient Multimodal Detection Transformer with Mamba-Driven Dual-Granularity Fusion and Frequency-Aware Modality Adapters · Cognitive Disentanglement for Referring Multi-Object Tracking · Seeing through Light and Darkness: Sensor-Physics Grounded Deblurring HDR NeRF from Single-Exposure Images and Events | 55.63 | 63 | 92% |
| 25 | University of Oxford FDIF: Formula-Driven Supervised Learning with Implicit Functions for 3D Medical Image Segmentation · Video Reality Test: Can AI-Generated ASMR Videos fool VLMs and Humans? · Societal Alignment Frameworks Can Improve LLM Alignment | 55.58 | 84 | 91% |
| 26 | City University of Hong Kong EHPE: Skeleton Cues-based Gaussian Coordinate Encoding for Efficient Human Pose Estimation · Approaching Low-Cost Cardiac Intelligence with Semi-Supervised Knowledge Distillation · UAV-enabled Computing Power Networks: Design and Performance Analysis under Energy Constraints | 55.58 | 70 | 92% |
| 27 | Southeast University Structure-Aware Multimodal LLM Framework for Trustworthy Near-Field Beam Prediction · Cognitive Disentanglement for Referring Multi-Object Tracking · Towards Perfection: Building Inter-component Mutual Correction for Retinex-based Low-light Image Enhancement | 55.58 | 67 | 92% |
| 28 | Purdue University Decoding Predictive Inference in Visual Language Processing via Spatiotemporal Neural Coherence · HccePose(BF): Predicting Front & Back Surfaces to Construct Ultra-Dense 2D-3D Correspondences for Pose Estimation · Align When They Want, Complement When They Need! Human-Centered Ensembles for Adaptive Human-AI Collaboration | 55.58 | 57 | 92% |
| 29 | University of Cambridge Efficient Camera-Controlled Video Generation of Static Scenes via Sparse Diffusion and 3D Rendering · Societal Alignment Frameworks Can Improve LLM Alignment · Dual role of core electrons in electronic friction | 55.46 | 88 | 92% |
| 30 | University of Electronic Science and Technology of China DRNet: All-in-One Image Restoration via Prior-Guided Dynamic Reparameterization · From Local Windows to Adaptive Candidates via Individualized Exploratory: Rethinking Attention for Image Super-Resolution · Simulation of complex DNA damage enhancement and biological effect validation for Proton-CAT | 55.46 | 67 | 92% |
| 31 | Tongji University Bridging Cognitive Neuroscience and Graph Intelligence: Hippocampus-Inspired Multi-View Hypergraph Learning for Web Finance Fraud · Beyond the Black Box: A Cognitive Architecture for Explainable and Aligned AI · Symplectic connection third-order Hall effect in a room-temperature ferromagnet | 55.46 | 61 | 91% |
| 32 | Georgia Institute of Technology FT-RAG: A Fine-grained Retrieval-Augmented Generation Framework for Complex Table Reasoning · Agentic Enterprise: AI-Centric User to User-Centric AI · TerraFormer: Automated Infrastructure-as-Code with LLMs Fine-Tuned via Policy-Guided Verifier Feedback | 55.46 | 57 | 94% |
| 33 | Xiamen University WarpGAN: Warping-Guided 3D GAN Inversion with Style-Based Novel View Inpainting · Endowing Molecular Language with Geometry Perception via Modality Compensation for High-Throughput Quantum Hamiltonian Prediction · Emergent surface resonance from charge density wave symmetry breaking in TiSe2 | 55.46 | 44 | 92% |
| 34 | Southern University of Science and Technology Empowering Autonomous Debugging Agents with Efficient Dynamic Analysis · MoNetV2: Enhanced Motion Network for Freehand 3D Ultrasound Reconstruction · Continuous Expert Assembly: Instance-Conditioned Low-Rank Residuals for All-in-One Image Restoration | 55.46 | 40 | 91% |
| 35 | New York University Solaris: Building a Multiplayer Video World Model in Minecraft · You Don’t Need Strong Assumptions: Visual Representation Learning via Temporal Differences · The Last Human-Written Paper: Agent-Native Research Artifacts | 55.41 | 85 | 94% |
| 36 | Wuhan University Universally Unfiltered and Unseen: Input-Agnostic Multimodal Jailbreaks against Text-to-Image Model Safeguards · Towards Perfection: Building Inter-component Mutual Correction for Retinex-based Low-light Image Enhancement · VOID: Defeating Unauthorized Mimicry in Latent Diffusion Models | 55.41 | 78 | 94% |
| 37 | University of Toronto The Last Human-Written Paper: Agent-Native Research Artifacts · CamDirector: Towards Long-Term Coherent Video Trajectory Editing · Societal Alignment Frameworks Can Improve LLM Alignment | 55.41 | 70 | 93% |
| 38 | Xi’an Jiaotong University Universally Unfiltered and Unseen: Input-Agnostic Multimodal Jailbreaks against Text-to-Image Model Safeguards · VOID: Defeating Unauthorized Mimicry in Latent Diffusion Models · Continuous Expert Assembly: Instance-Conditioned Low-Rank Residuals for All-in-One Image Restoration | 55.41 | 45 | 92% |
| 39 | Sichuan University MoGen: A Unified Collaborative Framework for Controllable Multi-Object Image Generation · LEARNING DOMAIN-AWARE TASK PROMPT REPRESENTATIONS FOR MULTI-DOMAIN ALL-IN-ONE IMAGE RESTORATION · SLER-IR: Spherical Layer-wise Expert Routing for All-in-One Image Restoration | 55.41 | 38 | 92% |
| 40 | Northeastern University Creativity as a Human Right: Design Considerations for Computational Creativity Systems · Shifting Adaptation from Weight Space to Memory Space: A Memory-Augmented Agent for Medical Image Segmentation · SimpleCall: A Lightweight Image Restoration Agent in Label-Free Environments with MLLM Perceptual Feedback | 55.35 | 69 | 92% |
| 41 | Johns Hopkins University RadThinking: A Dataset for Longitudinal Clinical Reasoning in Radiology · Counterfactual World Models via Digital Twin-conditioned Video Diffusion · Transformers Learn the Optimal DDPM Denoiser for Multi-Token GMMs | 55.23 | 73 | 93% |
| 42 | Shenzhen University LoBoFit: Flexible Garment Refitting via Local Bone Mapping Blending · MoNetV2: Enhanced Motion Network for Freehand 3D Ultrasound Reconstruction · DEPTH-GUIDED METRIC-AWARE TEMPORAL CONSISTENCY FOR MONOCULAR VIDEO HUMAN MESH RECOVERY | 55.18 | 49 | 94% |
| 43 | Northwestern Polytechnical University Toward Real-world Infrared Image Super-Resolution: A Unified Autoregressive Framework and Benchmark Dataset · TPGDiff: Hierarchical Triple-Prior Guided Diffusion for Image Restoration · UIR-LoRA: Achieving Universal Image Restoration through Multiple Low-Rank Adaptation | 55.18 | 49 | 94% |
| 44 | Princeton University AN IRREDUCIBLE REAL PROJECTIVE PLANE IN THE 4-SPHERE · DAG-Math: Graph-Guided Mathematical Reasoning in LLMs · Motion Attribution for Video Generation | 55.18 | 48 | 94% |
| 45 | Yale University Open-Ended Goal Inference through Actions and Language for Human-Robot Collaboration · The Last Human-Written Paper: Agent-Native Research Artifacts · Is Your Conditional Diffusion Model Actually Denoising? | 55.12 | 52 | 95% |
| 46 | Harvard University Beyond Anthropomorphism: a Spectrum of Interface Metaphors for LLMs · The Last Human-Written Paper: Agent-Native Research Artifacts · Exactly solvable non-unitary conformal interfaces in unitary CFTs | 55.12 | 51 | 96% |
| 47 | Cornell University Tracking and Understanding Object Transformations · The Last Human-Written Paper: Agent-Native Research Artifacts · Bayesian hypergraph inference from scarce and noisy dynamical observations | 55.06 | 74 | 96% |
| 48 | University of Illinois Urbana-Champaign CacheFlow: Efficient LLM Serving with 3D-Parallel KV Cache Restoration · MR. Video: “MapReduce” is the Principle for Long Video Understanding · LLM Alignment as Retriever Optimization: An Information Retrieval Perspective | 55.01 | 65 | 92% |
| 49 | University of Washington Neural Robot Dynamics · The Last Human-Written Paper: Agent-Native Research Artifacts · TomoGraphView: 3D Medical Image Classification with Omnidirectional Slice Representations and Graph Neural Networks | 55.01 | 62 | 96% |
| 50 | Technical University of Munich A New Perspective On AI Safety Through Control Theory Methodologies · Rewarding Doubt: A Reinforcement Learning Approach to Calibrated Confidence Expression of Large Language Models · TomoGraphView: 3D Medical Image Classification with Omnidirectional Slice Representations and Graph Neural Networks | 54.95 | 86 | 95% |
排名用于发现值得进一步核查的论文，不替代同行评审。评分会随新增引用、复现、实验验证、新闻评价和落地证据持续更新。
PaperMiner MCP已开放结构化论文反馈。可提交复现成功或失败、代码可运行或失效、Benchmark纠错、链接纠错和部署结果。
MCP接口：submit_paper_feedback、get_paper_community_feedback
权限：有效PaperMiner账号获配的个人MCP密钥可提交；公共代理和共享只读密钥只能查询。每个账号每天最多提交5条新反馈。