Paperglide public library — simplified academic papers Browse every academic paper Paperglide has simplified. Each page offers a TL;DR, semantic sections, and a link to the original source. Browse by topic cs.AI paperscs.LG paperscs.CL paperscs.CV paperscs.NE papersstat.ML papers All simplified papers Looped World Models (LoopWM)Geometric Action Model for Robot Policy LearningDOPD: Dual On-Policy DistillationFastContext: Training Efficient Repository Explorer for Coding AgentsScaling Laws for Language ModelsDeepSeek-V4: Towards Highly Efficient Million-Token Context IntelligenceWaypoint-OPD: Transferring RL-Induced Policy Shifts via Gated On-Policy DistillationOmniCompass: A Holistic Benchmark for Omni-Modal Video Grounding and UnderstandingStreamEval: Evaluating Interactive Streaming Video Generation ModelsOpenGCUA: Toward a Fully Open Stack for Generalist Computer UseHexaShell: A Thin Harness for Exploring the Limits of VLM Agents as Robot Coding AgentsPlayer-1: Learning to Play from Synthetic ExperienceFDE-Gym: The Last Mile of AI DeliveryHas GPT-6-Astra Achieved Embodied Intelligence?Jeva: System One Models from First PrinciplesReading, Writing, Sharing, and Interacting with the World through VideoRealCompanion: Benchmarking Human Understanding from Reasoning over Longitudinal Real-World ConversationsBeyond Domain-Routed Supervision: Cross-Teacher Guarding for Multi-Teacher On-Policy DistillationScaling Parameter-Efficient Fine-Tuning for Brain EncodingOR-License: Evaluating LLM Agents as Real-World Operations Research PractitionersFocused Omni-Modal GUI AgentsWeVisOne-VL: Unified Visual Perception from Objects to PixelsThinking Inertia: LLMs Keep Thinking When Told Not ToBizBench: Unveiling Hidden Challenges in Deploying Long-Horizon Agents for Real-World BusinessAgentSWE: Can Coding Agents Build the Agent You Actually Want?VideoJev: System One Model for Video Understanding and InteractionRoTA-Bench: Evaluating 3D Spatial Reasoning through Interactive Projection AlignmentAgent-Reflex: Teaching Language Agents to Act ReflectivelySelf-Evolving Agent HarnessMatrix-Game 4.0: Harnessing Video Models for Real-Time Playable Game WorldsRopedia W0: Scaling Loco-Manipulation World Models across Embodiments with Human ExperienceLEV: Thinking Fast and Deep with Looped Decision ModelsMaking MLLM Latents Matter: Factor-Aware Predictive Learning for Cinematic Video GenerationRethinking Cross-Tokenizer On-Policy Distillation: From Alignment Coverage to Supervision ReliabilityEgoXperience: Learning World-Grounded Egocentric Representations from Human ExperienceMAI-Gym: An Open All-in-One Platform for GUI Agent Evaluation and RL Training across Mobile and Computer EnvironmentsDeep Search in Real-World Web EnvironmentsOneStreamer: Unifying Perception, Memory, and Proactive Response in Streaming Video InteractionPAR: Progress-Aligned Recurrence for Length Generalization in State TrackingMJev-Spatial: System One Spatial Intelligence via Native Attention MatchingLearning to Think and Explore in 3D WorldHoli-World: Building Holistic World Intelligence from Human-Ego ExperienceLooking Ahead: Modeling Future Evolution in Autoregressive Video GenerationScaling Agentic Data for Long-Horizon Terminal IntelligenceAgentGarten: Code Worlds for Evolving AgentsDONIC: Generalist Dexterous Bimanual Tracking Control from Human DemonstrationsReflex: Visually Grounded Reactive Humanoid ControlLearning Where to Split: Dynamic Band Splitting for Audio Source SeparationSpatialScout: Can VLM Agents Actively Explore and Map 3D Scenes?Vidu S2: Real-Time Interactive, Editable, and Spatial Video GenerationCoEvo-RSI: When Harnesses Evolve, Which Models Should Follow?Spectral-Spatial InterpretationCache Engineering for AgentsKandinsky 6.0 Video: Foundation Models for Synchronized Video and Audio GenerationInfiniSplat v2: Analyzing and Improving Feed-Forward 3DGS from Any ViewsLiveMACE: Process-Aware Evaluation of LLM Agent Capabilities in Evolving MarketsUrbex: Agentic Spatial Grounding for City-Scale 3D ScenesTokenRouter: Efficient Serving System for Token-Level LLM RoutingLearn2Play Bench: How Well Do LLM Agents Learn from Experience in Unfamiliar Environments?Learning from Fleet Experience: A Collective Learning Paradigm for RobotsMemT-1.0: Towards Memory-Native EvolutionBehavioral Compositionality of Sparse Autoencoder RepresentationsPack Light, CarryOn: Building Object-Oriented Context for Long-Horizon AgentsRight Screen, Wrong Transition: World Models as Verifiers for GUI AgentsStudentSim: Training LLM-based Student Simulators3DCodeVerse: Exploring 3D Worlds as CodeRaven: The Harness of Harnesses for Composable Agentic IntelligenceFrom Traces to Agentic Worlds: Agentic Language World Models for Interactive Environment SimulationOpenVLHarness: Real World Visual Agentic Tool-Use with Multimodal MemoryTev: Scaling Long-Horizon Agentic Environments with Dense Progress RewardsDoc2Harness: Benchmarking LLMs on Long-Horizon Harness Engineering of Coding AgentsS-Agent 2.0: Spatial Reasoning with a General Coding AgentLong-WAM: Scaling the Context of World-Action ModelsPredictive Credit: Measuring What Scientific Explanations Add to Experimental ForecastsnanoMuse: An Open-Source Personal Agent for Every Device You OwnHarness^2: Recursive Agent Harnessing for an Open WorldAtria Dawn: The Dawn of Agentic SuperintelligenceScaling Trajectories for Complex Tasks through Recursive Self-RewriteSoftwareGym: Scaling Computer-Use Agent Training with Diverse Synthesized Software EnvironmentsTRACE: Rollout-Guided Quantization-Aware Training for FP4 Reinforcement Learning of MoE Language ModelsScaleNav: Scalable Internet Video Pretraining for Language-Conditioned Local NavigationRecursive Game Creator: An Agentic Product-Level Experience-Oriented Game HarnessGreenWAM: A World Foundation Model for RoboticsNeoHorse-1: Towards Recursive Self-Improvement via Agentic Post-Training with Routing HarnessAesCode: Aesthetic Code Generation with Decoupled Cross-Modal RewardsSpecMRL: Redesign Matryoshka Representation Learning within Parametric SpectrumVIS-Ground: Video Interactive Storytelling with Contextual GroundingSCOPE: Semantic Bounded Credit Optimization for Prompt Engineering in Text-to-image GenerationSpeculative Forcing for Efficient Autoregressive Video GenerationNative Action-Prior Learning from Videos for World Action ModelsWorld Action Modeling with Progressive Visual PlanningMore Methods, Better Agents? A Reality Check on Self-Distillation in Agentic Reinforcement LearningRecovering Reliable Supervision via Critique-Guided On-Policy DistillationPremiseCheck: Efficient Fact-Checking of Human and AI Peer ReviewsQwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous DrivingUltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image GenerationTPPO: Tracking-Value PPO For Language Model ReasoningGRACE: Generation-aware latent compression for efficient video generationWorkRover: Scaling Long-Horizon Cowork Agent Training Via Recursive Environment EvolutionEditing Images into Reusable Layers via Compositional Edit-Prior DistillationDuoMatching: Joint-Marginal Distribution Matching for Few-Step Video GenerationMaLiang-Harness: A Programmable Path to Image and Video GenerationBeyond Memory: Harnessing Long-Horizon Agents with Explicit Belief StatesSpeculative Sampling in Interactive Video GenerationHierarchical Continuous Diffusion Language ModelsSharpening Tax in Post-TrainingSelf-Retrospection Distillation: Turning Post-hoc Experiences into Prior ForesightDecepEval: A Benchmark for Evaluating Deception in LLM AgentsQuestioning the Questions: Sustaining Self-Evolution in Reasoning ModelsLatentLoop: Scaling Looped MoE Models with Asymmetric Compute AllocationWorld Observer: Joint Actor-Observer Generation for Persistent World ModelingSuperNav: An Agentic Navigation System for Any Task in Any SceneCortex: Consolidating Context into Parametric MemorySage-A1.5: From Long-Horizon Execution to Scientific Exploration with Capability CompositionAgent World Model in the Wild: Computer-Use Agent Environment SynthesisZGCM-1: A Fully Open and Extremely Efficient Foundation Model for Math and Agentic SearchEmbodiedCLBench: Evaluating Continual Learning for Self-Evolving Embodied AgentsContinual Learning Mechanisms Compose for Long-Horizon MemorizationALoDLM: Adaptively Looped Diffusion Language ModelsVLA-Rolling: Fewer Denoising Steps, Better Control via Adaptive Action ReuseROWBench: Do Video Models Render What the Program Specifies?MiMo-V2.6: Scaling Reinforcement Learning Towards Self-ImprovementCompile by Training: Turning Natural-Language Specifications into Local Neural FunctionsDream-RSI: Recursive Self-Improvement through Evolving WorldsVisionWeave: Weaving Elastic Visual Representations as a Native Capability of MLLMsEach Reward to Its Relevant Tokens: Reward-Specific Token-Level Credit Assignment for Multi-Reward RLLatent-MOPD: Latent Multi-Teacher On-Policy DistillationIQuest-Q1: Advancing Foundation Capabilities for Agentic CLI SystemsPost-Training Leaves Behavioral Shadows on Unrelated DecisionsActionOPD: Distilling Decisions, Not TokensVLA-Δ: Learning Physical State Transition for Vision-Language-Action ModelsMosaic: Multi-Session Agentic Reinforcement LearningGroundingPI: A Grounding Foundation Model towards Physical Intelligence with Visual PrimitivesSGF+: Decoupling Gradient Flows for Autoregressive Video GenerationScaling in Games: Continued Pre-Training For Embodied Agents In Diverse Virtual WorldsHyperBrowseComp: A Multilingual and Multimodal Stress Test for Web-Browsing AgentsCommerceGen-Studio: An End-to-End Framework for Commercial-Grade Main-Image GenerationUniEvo-VL: An On-policy Self-Distillation Training Recipe for Multimodal Model Self-improvementActiveSaddler: Automated Curriculum Learning for Agent Harness OptimizationScaling and Distilling Text Embeddings for Better DiffusibilityPrismaCoder: Pushing the Boundaries of Open Multimodal Code Intelligence through Decomposed Rubric SynthesisThe Teacher Is a Direction, Not a Destination: Extrapolating RL-Induced Representation Residuals in On-Policy DistillationMake Sparse Rewards Count: Density-Aware Reward Aggregation for Multi-Reward RLAutoGUIWorld: Image Generators as Visual World Models for GUI AgentProgramGym: Scaling Training Environments for Long-Horizon Program Reproduction AgentsToward Recursive Self-Improvement in Physical AIHarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?UserArena: Benchmarking Interactive User Simulation for Long-Horizon Shopping AgentsTerminal-Universe: Turning Agent Trajectories into Scalable Terminal EnvironmentsScaling Properties of Same-Family On-Policy DistillationSolarRenderer: Few-Step Generative Reference RenderingSeeing, Maintaining, and Learning: The Evolution of Multimodal Agent MemoryFewer Tokens, Better Action: GPT-6 Astra Robot Agents with 14% Higher Success Rate but 65% Fewer TokensYuE2: Unifying Symbolic and Audio Music Generation at Frontier QualityRethinking Latent Visual Reasoning: Grounding Latent Reasoning in Visual EvidenceVideo Generation Models: A Survey of Post-Training and AlignmentMulti-Agent Egocentric World Model with Fine-Grained Embodied InteractionSelf-Evolving Coding Agents via Harness-Policy OptimizationLLaDA-Image: Building Strong Image Generators with Fully Open Training RecipesInterEvolve: Test-Time Evolution of Reward Programs for Humanoid Loco-ManipulationIntern Lumina U2: A Multi-Codebook Diffusion Large Language Model for Omni-Visual Understanding and Image GenerationLearning the 4D World through Point Trajectory ForecastingDecoding Looped Transformers Better for (Almost) FreeContinually Post-training Small Models via Synthetic Self-distillationFalse Frontiers: Diagnosing and Mitigating Co-Cheating in Self-Evolving Search AgentsTraining Object Permanence in World ModelsWork2Skill: Capturing Human GUI Experience for Reusable Agent Skill InductionLensErase: A Foundation Model for Erase AnythingDeep Fusion Memory: A Chunk-Level Parametric Memory for Large Language ModelsSepGen: Multi-Stem Audio-Video Separation and Generation in a Single ModelHiSTEVE 1.0: Harnessing Human-Interface VLA Agents for Open-World GamesAuK Technical Report: An Open-Source Foundational Model for Speech Generation and EditingOmni-IO Skills: Harnessing Your Agent Omni-NativeNCP-ArchPreview Technical Report: Moving towards Latent Space Language Models through Next Concept PredictionEVISKILL: Grounding Skill Evolution in Replayable EvidenceU-Space: Uncovering When and Why Uncertainty Arises in Language ModelsREMORY: Learning Residual Memory for Context CompactionMC-Sparse: Deconstructing and Closing the Dense-Sparse Attention Gap in Diffusion TransformersAutoSFT: An Agentic System for Automated Data Research in Supervised Fine-TuningWhat Generation Knows: Unlocking Fine-Grained Understanding in Unified Multimodal ModelsPhysBrain 1.5: From Vision-Language Models to Physical Foundation ModelsDeltaStreamer: Towards Human-Like Agentic Streaming Video UnderstandingFrom Evidence to Action: How Tool-Using Agents FailRynnUnify: Towards Unified Embodied BrainRSIHub: A Composable Framework for Agent-Evolution MethodsRRSI: Regularized Recursive Self-Improvement of Agent HarnessesIn-Distribution Forcing for Long Video Generation at Test TimeDreamTrue: Action-Faithful Robot World Model with Counterfactual Post-TrainingMemento 3: Model-Based Recursive Self-Improvement through Reflective RulebooksTestPrism: Rethinking Test Evaluation Beyond a Single ReferenceMulti-Stream Attention ResidualsMemadapter: Counterfactual Adaptation Against Memory-induced SycophancyCan AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent CompetitionRandom Attention: Rethinking KV Cache Eviction for Efficient ReasoningBenchmark Radar: A Living Database and Search Engine for AI Benchmarks and EvaluationBeyond Spatio-Temporal Priors: A Generalizable Approach for Dense Correspondence MatchingLMBuild: Evaluating LLM Agents for Generating Buildable and Functional StructuresOuroWorld: Bringing Any 3D World Alive as Diverse, Endlessly Looping 3D CinemagraphsCan World and Action Modeling Benefit Each Other under Unified Causal Training?Orbia: Do Generated Videos Form a Consistent 3D World?