KAIST AI · 72 records

dateidtitleauthorsvenue↑
2026-11-202611.22666VSpark: Speculative Decoding for Autoregressive Vision-Language-Action ModelsSuho Yoo et al.arXiv5
2026-11-192607.21179Out of Sight, Still in Mind: Token Compression for Omni-LLMsSuho Yoo et al.arXiv3
2026-11-162611.18529eDIAR: Training-Free Open-Speaker Streaming Diarization for Long Meetings and Large Archives via e-meansKyudan Jung et al.arXiv9
2026-11-132611.16409Experience-Handoff Reinforcement Learning for Recursive Self-ImprovementDoyoung Kim et al.arXiv4
2026-11-102611.14320Don't Just Trust Your Memory: Evidence-Linked Memory Reduces Memory-Induced Degradation in LLM AgentsYukyeong Lee et al.arXiv61
2026-11-062611.04527Future Representation Distillation for Imagination-free World Action ModelsJaehyeong Jo et al.arXiv9
2026-11-052611.04391Human-to-Robot Transfer via Bridging Representation and Action Space in VLAsJunhyeong Park et al.arXiv14
2026-11-042611.05163RoboTryOn: Synthesizing Dexterous Robot Demonstrations from Human Videos via Virtual Try-OnSeungyong Lee et al.arXiv6
2026-11-032611.05377DreamX: Video Diffusion Model is a Universal Latent Action LearnerJisoo Kim et al.arXiv54
2026-11-032611.05193Factorized MLLM Embeddings for Composed Image RetrievalWonJun Moon et al.arXiv47
2026-11-022610.37393Learning Quality-Controllable Robot Policies with Iterative Hindsight RelabelingSinjae Kang et al.arXiv2
2026-10-302610.42149Latent Reasoning in Continuous Space for Unified Multimodal ModelsByungwoo Jeon et al.arXiv37
2026-10-292610.40614L3G-World: Real-Time Infinite World Generation with Latent 3D Gaussian MemoryJinhyeok Choi et al.arXiv39
2026-10-282610.37386Channel Selection for Mixture-of-Experts Decoding Needs No Calibration DataGeunhyoung Kim et al.arXiv17
2026-10-272610.36671Look Only When Needed: Certificate-Guided Visual Routing for Efficient Diffusion VLMsDonghyeon Cho et al.arXiv100
2026-10-272610.32195Principia: Test-Time Hypotheses to Principle Construction for Scientific DiscoveryDongki Kim et al.arXiv40
2026-10-262610.35684OmniShowcase: Omnidirectional Subject-Consistent Video Generation from Multi-View ImagesJunyeong Ahn et al.arXiv120
2026-10-212610.30320Self-Distilling Future Video Representations for World Action ModelsHuiwon Jang et al.arXiv47
2026-10-202610.27954Lost in Recovery: Demystifying Recovery Behavior of Computer-Use Agents via Goal-Oriented Action ReplayByeonghun Bang et al.arXiv33
2026-10-152610.14448KV4-RL: Unlocking Faster Asynchronous Rollouts for RLVR with INT4 KV Cache QuantizationSihwan Park et al.arXiv40
2026-10-142610.14228Latent World Action Model for GUI AgentsSeungjun Yu et al.arXiv21
2026-10-092610.12299Multi-Agent Egocentric World Model with Fine-Grained Embodied InteractionDahyun Chung et al.arXiv61
2026-10-082610.10524GRACE: Generation-aware latent compression for efficient video generationJiyoung Kim et al.arXiv73
2026-10-082610.10539Tetris3D: 3D Scene Generation With Objects That Fit TogetherJaeyeong Kim et al.arXiv48
2026-10-082610.10444RunningTab: Direct Workspace Interaction with Environment-Side TabsJinheon Baek et al.arXiv43
2026-10-082610.10437Q-Learning with Scalar Adjoint MatchingYonghoon Dong et al.arXiv25
2026-10-082610.02832FastOPD: On-Policy Distillation for Lightweight VLA DeploymentYoojin Oh et al.arXiv2
2026-10-082609.38562LongTake: Learning to Sustain Dynamics in Long-Horizon Video GenerationByoungwoo Park et al.arXiv1
2026-10-072610.02858Harness-Aware Distillation for Small Language Model AgentsMoonseok Choi et al.arXiv15
2026-10-062610.04929RobotUse: Allocating Computation, Context, and DecisionsJunhoo Lee et al.arXiv17
2026-10-022610.02162World Observer: Joint Actor-Observer Generation for Persistent World ModelingHyunwook Choi et al.arXiv63
2026-10-022609.34677Learning What to Recall: Adaptive Multi-Cue Episodic Memory for World ModelsBeomsu Kim et al.arXiv6
2026-10-012609.38177Imagine3D-LLM: Teaching MLLMs to Imagine 3D Scenes Before AnsweringJaewoo Jung et al.arXiv50
2026-10-012609.37915Overcoming Scaling Limits in On-Policy Self-Distillation for LLM ReasoningMd. Ismail Hossain et al.arXiv7
2026-09-302609.37868Learning Beyond What You Sample: Off-Policy-Aware Cross-Model Trajectory Exchange for RLVRDoohyuk Jang et al.arXiv49
2026-09-292609.33781Surprising Success, Repeated Failure: Entropy-Guided Credit Assignment for Exploration in LLM ReasoningWoongyeong Yeo et al.arXiv44
2026-09-292609.34327Knowing When Thinking Is Not Enough: Teaching Small Reasoning Models to Reason Beyond Their Parametric KnowledgeChanuk Lee et al.arXiv39
2026-09-172609.17632EvolveTrade: Experience-Driven Policy Refinement for Self-Evolving LLM Trading AgentsSehee Kim et al.arXiv25
2026-09-142608.30597PLC-DPO: Posterior Label Correction in Noisy and Ambiguous Preference OptimizationBoryeong Cho et al.arXiv27
2026-09-112609.07064SpatialBlock: Enhancing Spatial Intelligence in LVLMs via Synthetic Block-Stacking ProblemSoohyun Ryu et al.arXiv62
2026-09-092609.08798Eliciting Weak-to-Strong Generalization with On-Policy Reverse DistillationYoungrok Park et al.arXiv68
2026-09-032609.02737Language Models Can Control Their Own AttentionNamgyu Ho et al.arXiv77
2026-09-012608.29310Super Library Agent: Joint Generation and Maintenance of Multiple Applications Beyond the Single CodebaseDaegyu Sung et al.arXiv28
2026-08-312608.26582J-Zero: Unified Challenger--Solver--Judge Co-Evolution from Zero DataGyouk Chu et al.arXiv40
2026-08-212608.19861PolicyGuide: From Guarding One Action to Guiding the Whole Workflow for Policy-Compliant LLM AgentsSeongjae Kang et al.arXiv5
2026-08-132608.11616MBA: Multimodal Benchmark and Agents for Real-World Business IdeationHojun Choi et al.arXiv7
2026-08-112608.07169Agent Memory Distillation: Empowering Small LLM Agents with Hierarchical Teacher MemoryTaeil Kim et al.arXiv45
2026-07-292607.25565ReDesign: Recovering Editable Design Structures from Images via Agentic DecompositionJooyeol Yun et al.arXiv55
2026-07-202607.11498See like a Robot: Robot-Centric Pointmaps for Vision-Language-Action ModelsByungkun Lee et al.arXiv5
2026-07-082607.04412LLM-as-a-Tutor: Policy-Aware Prompt Adaptation for Non-Verifiable RLYujin Kim et al.arXiv36
2026-07-082606.313293D HAMSTER: Bridging Planning and Control in Hierarchical Vision Language Action Models through 3D Trajectory GuidanceDongyoon Hwang et al.arXiv7
2026-06-302606.29225PolicyGuard: A Dialogue-Grounded Sub-Agent Verifier for Policy Adherence in LLM AgentsSeongjae Kang et al.arXiv6
2026-06-252606.26087MVTrack4Gen: Multi-View Point Tracking as Geometric Supervision for 4D Video GenerationJoungBin Lee et al.arXiv35
2026-06-232606.23557Dense Reward for Multi-View 3D Reasoning with Global Maps and Local ViewsJiho Choi et al.arXiv7
2026-06-172606.09030TRIAGE: Dialectical LLM Reasoning for Explainable Risk Prediction on Irregularly Sampled Medical Time SeriesHyeongwon Jang et al.arXiv27
2026-06-162606.16281Who Should Lead Decoding Now? Tracking Reliable Trajectories for Ensembling Masked Diffusion Language ModelsHeecheol Yun et al.arXiv32
2026-06-102606.11180Lip Forcing: Few-Step Autoregressive Diffusion for Real-time Lip SynchronizationPaul Hyunbin Cho et al.arXiv36
2026-06-052606.04743TIDE: Proactive Multi-Problem Discovery via Template-Guided IterationSoyeong Jeong et al.arXiv47
2026-06-052605.27079Trust Region Q Adjoint MatchingYonghoon Dong et al.arXiv24
2026-06-032606.02578Mitigating Perceptual Judgment Bias in Multimodal LLM-as-a-Judge via Perceptual Perturbation and Reward ModelingSeojeong Park et al.arXiv4
2026-05-292605.29250OmniRetrieval: Unified Retrieval across Heterogeneous Knowledge SourcesJinheon Baek et al.arXiv77
2026-05-292605.27355Alignment Tampering: How Reinforcement Learning from Human Feedback Is Exploited to Optimize Misaligned BiasesDongyoon Hahm et al.arXiv2
2026-05-282605.28775Learn from Weaknesses: Automated Domain Specialization for Small Computer-Use AgentsSuji Kim et al.arXiv34
2026-05-272605.26230Geometry-Aware Representation Denoising for Robust Multi-view 3D ReconstructionJin Hyeon Kim et al.arXiv37
2026-05-252605.17873HINT-SD: Targeted Hindsight Self-Distillation for Long-Horizon AgentsWoongyeng Yeo et al.arXiv10
2026-05-222605.22718WorldKV: Efficient World Memory with World Retrieval and CompressionJung Yi et al.arXiv39
2026-05-222605.20910FlowLong: Inference-time Long Video Generation via Manifold-constrained Tweedie MatchingJangho Park et al.arXiv24
2026-05-212605.20258It Takes Two: Complementary Self-Distillation for Contextual Integrity in LLMsSangwoo Park et al.arXiv29
2026-05-182605.15726Nudging Beyond the Comfort Zone: Efficient Strategy-Guided Exploration for RLVRChanuk Lee et al.arXiv20
2026-05-152605.13880PREPING: Building Agent Memory without TasksYumin Choi et al.arXiv28
2026-05-132605.12477MEME: Multi-entity & Evolving Memory EvaluationSeokwon Jung et al.arXiv6
2026-05-122605.08735CollabVR: Collaborative Video Reasoning with Vision-Language and Video Generation ModelsJoowon Kim et al.arXiv71