Shanghai AI Laboratory · 25 records

dateidtitleauthorsvenue↑
2026-11-182611.23144π^4: A Foundation Model for 4D Scene Geometry and MotionYifan Wang et al.arXiv13
2026-11-182611.23140GameRealm: Grounding Visual Learning in Interactive GamesWeiguang Zhao et al.arXiv1
2026-11-092611.12762Sage-Taiyi: A Metacognitive Architecture for Scientific DiscoveryWeida Wang et al.arXiv54
2026-10-272610.35518S3AC: Stabilized, Selective, and Scalable Actor-Critic for LLM ReasoningShuo Shen et al.arXiv33
2026-10-272610.33005OpenESM: Energy-Steered Models Are Scalable Foundation LearnersYuan Pu et al.arXiv6
2026-10-192610.21767From Research Goals to Scientific Closure: Evaluating Agents Through Evidence-Supported Physics ResearchYijia Mao et al.arXiv47
2026-10-162610.20818Intern Lumina U2: A Multi-Codebook Diffusion Large Language Model for Omni-Visual Understanding and Image GenerationQian Yu et al.arXiv50
2026-10-162610.19397SERO: Learning from Executable Oracle Continuations for Long-Horizon Agentic Reinforcement LearningDingyi Rong et al.arXiv8
2026-10-052609.33463Rethinking Token Reweighting for SFT: Suppress, Reverse, and Extrapolate Learned FeaturesCunchun Li et al.arXiv2
2026-09-302609.37089Real2Gym: Building Gyms from Videos, Bringing Skills to RobotsKerui Ren et al.arXiv6
2026-09-292609.35743InfiniHand: Streaming World-Space Hand Motion Estimation from Egocentric VideoKerui Ren et al.arXiv9
2026-09-252609.29816AV-GRPO: Modality-Anchored Decoupling Diffusion Reinforcement Learning for Joint Audio-Video GenerationZhiyu Xu et al.arXiv9
2026-09-172609.18708Rethinking Critic Learning in PPO: Understanding and Mitigating Value FlatteningYizhuo Li et al.arXiv63
2026-09-112609.10715NCP-ArchPreview Technical Report: Moving towards Latent Space Language Models through Next Concept PredictionNCP Team et al.arXiv176
2026-09-102609.08149SWE-Bench Pro Verified: A Reliable Benchmark for Software Engineering AgentsPujun Zheng et al.arXiv10
2026-09-022609.00092Safin-1: Safety from Within through Memory-Native State EvolutionMing Zhang et al.arXiv20
2026-09-022608.28612InternReviewer & InternAdvocate: Objective Reward and Evaluation for Agentic Reinforcement Learning in Peer Review and RebuttalXuerui Su et al.arXiv6
2026-08-182608.14783MegaParts: Scaling Part-Aware 3D Object Generation to 300 Parts via Token-Efficient Autoregressive ModelingManwen Liao et al.arXiv20
2026-08-182608.15304Understanding Cognition-Induced Risks in Agentic AI SystemsGuanchu Wang et al.arXiv17
2026-08-172608.14277SimpleOPD: Simple Tokenizer-Agnostic On-Policy Distillation for Long-Context ReasoningHaonan He et al.arXiv36
2026-08-112608.07346A^2E : An End-to-End Agent Auditing EngineHaoning Wang et al.arXiv10
2026-08-052608.02713Quo Vadis, World Modeling?Yu Yang et al.arXiv34
2026-08-042608.01735DAPD: Dual-Anchored Policy DistillationJianyu Wu et al.arXiv35
2026-08-042608.02287SKT: Skill-Use Training at Scale via Verified Synthetic Data GenerationZelin Tan et al.arXiv25
2026-07-092607.07708Accurate, Interdisciplinary and Transparent Structure-property Understanding with Deep Native Structural ReasoningChen Tang et al.arXiv81