Peking University · 67 records

dateidtitleauthorsvenue↑
2026-11-192611.24830A Foundation Model to Measure and Predict Human BehaviourHaoran Ye et al.arXiv2
2026-11-182611.19857PanDWM: Panoramic 4D Driving World ModelsJiajun Sun et al.arXiv1
2026-11-172611.19771JEPA4R: Embodied JEPA World Models with 4D Point PlanningJiarui Dong et al.arXiv5
2026-11-13W4416236439Emergent long-tailed dynamics in spin iceChenyue Guo et al.Proceedings of the National Academy of Sciences
2026-11-032610.40293MAServeBench: Benchmarking Multi-Agent LLM Serving Systems with Hierarchical Collaboration WorkloadsShengxuan Qiu et al.arXiv9
2026-11-03W7133615810Helios: Real Real-Time Long Video Generation ModelShenghai Yuan et al.IEEE Transactions on Pattern Analysis and Machine Intelligence
2026-10-272610.37341CodeCap-Bench: Can Code Serve as a Good Video Caption?Jiafu Tang et al.arXiv52
2026-10-272610.33079Rethinking Video Storage for AI Training: Breaking Cross-Frame Dependencies for Efficient Random AccessZetian Song et al.arXiv12
2026-10-202609.22332AffordanceWAM: Affordance-Aware Joint World-Action Modeling for Robot ManipulationJiadi You et al.arXiv20
2026-10-192610.26659SpatialScout: Can VLM Agents Actively Explore and Map 3D Scenes?Zhaoxi Zhang et al.arXiv120
2026-10-162610.22240MMSearch-OPD: Easy Made Hard, Hard Made Learnable for Long-Horizon Multimodal Search AgentsHuanyao Zhang et al.arXiv6
2026-10-162610.06598SimForcing: Distilling Simulation Motion Priors into Real-Domain Robot World ModelsXiaodong Wang et al.arXiv2
2026-10-142610.17656Open-o3-Video-v2: Seek, Ground, and Reason with Verifiable Spatio-Temporal EvidenceJiahao Meng et al.arXiv24
2026-10-142602.05758LongR: Unleashing Long-Context Reasoning via Reinforcement Learning with Dense Utility RewardsQize Yu et al.arXiv21
2026-10-142610.15357UniForge3D: Joint 3D Reconstruction and Generation in a Shared Anchor-Guided SpaceWenjue Chen et al.arXiv4
2026-10-12W4415137648Antiferromagnetism and tightly bound Cooper pairs induced by kinetic frustrationYixin Zhang et al.Proceedings of the National Academy of Sciences
2026-10-082610.10047AdSpark: A Large-Scale Dataset and Benchmark for Product-Centric Advertisement Video GenerationZhifei Yang et al.arXiv7
2026-10-062610.05033Code2Games: Enabling Coding Agents for Gaming World GenerationWei Wu et al.arXiv5
2026-10-052610.02388Octrees as an Explicit 3D LanguageRan Dan et al.arXiv11
2026-10-022610.00360DexPolicy: Scheduled Exploration for Trajectory-Guided Dexterous ManipulationHaoyu Wang et al.arXiv2
2026-09-302609.35932Same Bytes, Different Authority: Reserved-Token Representations in Chat-Template Prompt InjectionYan Zhan et al.arXiv5
2026-09-292609.29233Post-Training Leaves Behavioral Shadows on Unrelated DecisionsZiyang Zhang et al.arXiv48
2026-09-292609.31948Duplex-MPE: Benchmarking Multi-Party Interaction in Full-Duplex DialogueChengqian Ma et al.arXiv43
2026-09-282609.18703RayOrch: Programming and Executing Lineage-Controlled Multi-Grain Dataflows for Foundation-Model Data PreparationXiaochen Ma et al.arXiv53
2026-09-252609.28811DeltaWAM: Delta World Action Models for Bimanual ManipulationHan Yan et al.arXiv17
2026-09-222609.23088OmniEdu: Open Foundation Models for Learning and TeachingHao Liang et al.arXiv69
2026-09-222609.24974Harness-Zero: Harness Distillation via Agent-as-HarnessHaoran Ye et al.arXiv34
2026-09-142609.06107DataFlex-RL: An Evaluation Platform for RLVR Data PoliciesHao Liang et al.arXiv66
2026-09-142609.06251MobileVLA-R1 2.0: RL-Enhanced Reasoning for Mobile Robot ControlTing Huang et al.arXiv6
2026-09-142609.08365ReMoMask-2: Latent Retrieval-Augmented Masked Motion GenerationYiran Wang et al.arXiv1
2026-09-102609.02771From Reweighting to Rewriting: Unlocking the Intervention Effects of Influential Samples in Training Data AttributionYuzhang Luo et al.arXiv4
2026-09-072609.04250Motion-Omni: End-to-End Joint Speech and Full-Body Motion for Spoken DialogueChengqian Ma et al.arXiv43
2026-08-262608.22274Length-Adaptive Decoding for Masked Diffusion Machine TranslationYan Zhan et al.arXiv3
2026-08-252608.23252The Laws of Context Allocation: Causal Measurement and Closed-Loop Orchestration in Generative SearchPeiyang Liu et al.arXiv4
2026-08-172608.00220Verifier-Induced Support Reshaping in On-Policy OptimizationShaohang Wei et al.arXiv6
2026-08-132608.11878ToolHazard: Scaling Adversarial Environments for Security Evaluation and Alignment of LLM-based AgentsYutao Mou et al.arXiv13
2026-08-052608.01127MiniWorld: Democratizing the Training of Video World Models from ScratchYian Zhao et al.arXiv16
2026-08-052608.03971UniWorld-Design: From Pixel Generation to Layer-Native DesignZongjian Li et al.arXiv13
2026-08-052608.03874ContinualSkillBench: Can LLM Agents Truly Evolve Their Capabilities?Tianyi Guan et al.arXiv10
2026-08-042608.01827DeepVoyager-VL: Incentivizing Vision-in-the-Loop Search for Long-Horizon Multimodal AgentsHuanyao Zhang et al.arXiv13
2026-08-042608.00574Relax Within, Balance Across: Geometry-Guided Load Balancing for Vision-Language Mixture-of-ExpertsZiang Wu et al.arXiv3
2026-07-312607.28022Flux-OPD: On-Policy Distillation with Evolving ContextsYuran Wang et al.arXiv45
2026-07-282607.24744Data Pyramid for Embodied ManipulationYifan Ye et al.arXiv36
2026-07-272607.20465DataPrep-Bench: Benchmarking LLMs as Training Data PreparatorsHao Liang et al.arXiv53
2026-07-242605.09635K12-KGraph: A Curriculum-Aligned Knowledge Graph for Benchmarking and Training Educational LLMsHao Liang et al.arXiv55
2026-07-222607.16617DataFlow-Harness: A Grounded Code-Agent Platform for Constructing Editable LLM Data PipelinesRunming He et al.arXiv87
2026-07-222607.18091SciForma: Structure-Faithful Generation of Scientific DiagramsYuxuan Luo et al.arXiv24
2026-07-132607.06374VaseMuseum: Digital Intelligent Museum for Ancient Greek PotteryJiazi Wang et al.arXiv4
2026-06-292606.27978Parallel Rollout Approximation for Pixel-Space Autoregressive Image GenerationJiayi Xu et al.arXiv5
2026-06-222606.17480GeneralVLA-2: Geometry-Aware Reconstruction and Governed Memory for Robot PlanningHaoyu Wang et al.arXiv3
2026-06-192606.15133DragMesh-2: Physically Plausible Dexterous Hand-Object Interaction with Articulated ObjectsTianshan Zhang et al.arXiv33
2026-06-172606.15142MotionVLA: Vision-Language-Action Model for Humanoid MotionNonghai Zhang et al.arXiv4
2026-06-082606.07433Watch, Remember, Reason: Human-View Video Understanding with MLLMsJiahao Meng et al.arXiv20
2026-06-052606.06042LoomVideo: Unifying Multimodal Inputs into Video Generation and EditingJianzong Wu et al.arXiv24
2026-06-052606.06155AffordanceVLA: A Vision-Language-Action Model Empowering Action Generation through Affordance-Aware UnderstandingQize Yu et al.arXiv18
2026-06-052606.03092The Shadow Price of Reasoning: Economic Perspective on Optimal Budget Allocation for LLMsXu Wan et al.arXiv7
2026-05-282605.28691OSP-Next: Efficient High-Quality Video Generation with Sparse Sequence Parallelism, HiF8 Quantization, and Reinforcement LearningYunyang Ge et al.arXiv21
2026-05-212605.14747Video2GUI: Synthesizing Large-Scale Interaction Trajectories for Generalized GUI Agent PretrainingWeimin Xiong et al.arXiv46
2026-05-202605.18263RT-Splatting: Joint Reflection-Transmission Modeling with Gaussian SplattingJi Shi et al.arXiv7
2026-05-192605.18287StableVLA: Towards Robust Vision-Language-Action Models without Extra DataYiyang Fu et al.arXiv14
2026-05-182605.15250GQLA: Group-Query Latent Attention for Hardware-Adaptive Large Language Model DecodingFanxu MengarXiv12
2026-05-152605.15186VGGT-Edit: Feed-forward Native 3D Scene Editing with Residual Field PredictionKaixin Zhu et al.arXiv24
2026-05-142605.11363PresentAgent-2: Towards Generalist Multimodal Presentation AgentsWei Wu et al.arXiv8
2026-05-132605.10365Agent-ValueBench: A Comprehensive Benchmark for Evaluating Agent ValuesHaonan Dong et al.arXiv9
2026-05-132605.11354Lite3R: A Model-Agnostic Framework for Efficient Feed-Forward 3D ReconstructionHaoyu Zhang et al.arXiv2
2026-05-112605.07363MISA: Mixture of Indexer Sparse Attention for Long-Context LLM InferenceRuijie Zhou et al.arXiv11
2026-05-042604.24026From Skill Text to Skill Structure: The Scheduling-Structural-Logical Representation for Agent SkillsQiliang Liang et al.arXiv21