Papers

New on arXiv

Mon 12 Jan 2026

370 papers

CRAFT: Fine-Grained Cost-Aware Expert Replication For Efficient Mixture-of-Experts ServingMixture-of-experts → Efficient MoE training and inference systems
Bridge-RAG: An Abstract Bridge Tree Based Retrieval Augmented Generation AlgorithmRetrieval-augmented generation → RAG pipeline design and evaluation
EdgeNav-QE: QLoRA Quantization and Dynamic Early Exit for LAM-based Navigation on Edge DevicesEfficient LLM inference and serving → Visual token pruning in multimodal models
A Safety-Constrained Reinforcement Learning Framework for Reliable Wireless AutonomyDeep reinforcement learning methods → Safe and constrained RL
G-MemLLM: Gated Latent Memory Augmentation for Long-Context Reasoning in Large Language ModelsLLM reasoning → Test-time compute and efficient thinking
CascadeMind at SemEval-2026 Task 4: A Hybrid Neuro-Symbolic Cascade for Narrative SimilarityLLM reasoning → Reasoning benchmarks and evaluation
Reasoning Stabilization Point: A Training-Time Signal for Stable Evidence and Shortcut RelianceLLM reasoning → Test-time compute and efficient thinking
SAGE: Tool-Augmented LLM Task Solving Strategies in Scalable Multi-Agent EnvironmentsLLM agents and tool use → General LLM agent frameworks and benchmarks
TAG-MoE: Task-Aware Gating for Unified Generative Mixture-of-ExpertsDiffusion models → Diffusion foundations and sampling +1
Semantic visually-guided acoustic highlighting with large vision-language modelsMultimodal LLMs / vision-language models → Video understanding with MLLMs
MemoBrain: Executive Memory as an Agentic Brain for ReasoningLLM reasoning → Reasoning for agents, planning and code +1
Semantic Gravity Wells: Why Negative Constraints BackfireMechanistic interpretability → Knowledge and representations inside LLMs
Triggering Chain-of-Thought via Latent Feature Interventions in Large Language ModelsLLM reasoning → Prompting and chain-of-thought reasoning +1
Forecast Aware Deep Reinforcement Learning for Efficient Electricity Load Scheduling in Dairy FarmsDeep reinforcement learning methods → RL for energy, networks and operations
Unifying Hamilton-Jacobi Reachability and Reinforcement LearningDeep reinforcement learning methods → RL theory and sample complexity
InfGraND: An Influence-Guided GNN-to-MLP Knowledge DistillationKnowledge distillation → Graph neural network distillation
Training Free Zero-Shot Visual Anomaly Localization via Diffusion InversionDiffusion models → Flow matching, time series and science
Representations of Text and Images Align From Layer OneMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Towards Verifiably Safe Tool Use for LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
TP-Blend: Textual-Prompt Attention Pairing for Precise Object-Style Blending in Diffusion ModelsDiffusion models → Image generation, text-to-image and editing
CASHEW: Stabilizing Multimodal Reasoning via Iterative Trajectory AggregationLLM reasoning → Multimodal reasoning +1
Reasoning over Precedents Alongside Statutes: Case-Augmented Deliberative Alignment for LLM SafetyLLM reasoning → Reasoning for agents, planning and code +2
VULCA-Bench: A Multicultural Vision-Language Benchmark for Evaluating Cultural UnderstandingMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Cross-Cultural Expert-Level Art Critique Evaluation with Vision-Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Cost and Accuracy of Long-Term Memory in Distributed Multi-Agent Systems Based on Large Language ModelsLLM agents and tool use → Agent memory systems
Executable Ontologies in Game Development: From Algorithmic Control to Semantic World ModelingWorld models → World models for language agents
A Human-Centric Pipeline for Aligning Large Language Models with Chinese Medical EthicsLLM alignment / RLHF / preference optimisation → Medical LLM alignment
Reinforcement Learning Methods for Neighborhood Selection in Local SearchDeep reinforcement learning methods → RL for games and combinatorial optimization
Coupled Diffusion-Encoder Models for Reconstruction of Flow FieldsDiffusion models → Flow matching, time series and science
Towards Specialized Generalists: A Multi-Task MoE-LoRA Framework for Domain-Specific LLM AdaptationMixture-of-experts → Medical and biomedical MoE
Revealing the Attention Floating Mechanism in Masked Diffusion ModelsDiffusion models → Discrete and text diffusion
KVzap: Fast, Adaptive, and Faithful KV Cache PruningEfficient LLM inference and serving → KV cache optimization
SecureCAI: Injection-Resilient LLM Assistants for Cybersecurity OperationsLLM safety and security → Prompt injection and agent security
Robotic Video World Models: A Survey of Applications, Research Challenges, Future DirectionsWorld models → Latent predictive world models
Failure-Aware RL: Reliable Offline-to-Online Reinforcement Learning with Self-Recovery for Real-World ManipulationOffline RL and imitation learning from data → Safe offline RL
More Images, More Problems? A Controlled Analysis of VLM Failure ModesMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Kinship Data Benchmark for Multi-hop ReasoningLLM reasoning → Reasoning benchmarks and evaluation
Benchmarking Small Language Models and Small Reasoning Language Models on System Log Severity ClassificationLLM reasoning → Reasoning for agents, planning and code
Beyond Single-Shot: Multi-step Tool Retrieval via Query PlanningLLM agents and tool use → Tool learning and function calling
Enhancing Self-Correction in Large Language Models through Multi-Perspective ReflectionLLM reasoning → Prompting and chain-of-thought reasoning
OS-Symphony: A Holistic Framework for Robust and Generalist Computer-Using AgentMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
Self-transcendence: Is External Feature Guidance Indispensable for Accelerating Diffusion Transformer Training?Diffusion models → Diffusion foundations and sampling
Video Evidence to Reasoning Efficient Video Understanding via Explicit Evidence GroundingLLM reasoning → Multimodal reasoning +1
Structure First, Reason Next: Enhancing a Large Language Model using Knowledge Graph for Numerical Reasoning in Financial DocumentsLLM reasoning → Reasoning for agents, planning and code
Seeing vs. Believing: Evaluating the Language Bias of Open-Source MLLMs in Counter-Intuitive ScenesMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Is Agentic RAG worth it? An experimental comparison of RAG approachesRetrieval-augmented generation → Agentic RAG, agents and memory +1
Exploring the Meta-level Reasoning of Large Language Models via a Tool-based Multi-hop Tabular Question Answering TaskLLM reasoning → Reasoning benchmarks and evaluation
Smooth Operator: Smooth Verifiable Reward Activates Spatial Reasoning Ability of Vision-Language ModelLLM reasoning → Multimodal reasoning +1
R3DPA: Leveraging 3D Representation Alignment and RGB Pretrained Priors for LiDAR Scene GenerationWorld models → World modeling from LiDAR generation
Adaptive Layer Selection for Layer-Wise Token Pruning in LLM InferenceEfficient LLM inference and serving → KV cache optimization
Reasoning Models Will Sometimes Lie About Their ReasoningLLM reasoning → Prompting and chain-of-thought reasoning
PlaM: Training-Free Plateau-Guided Model Merging for Better Visual Grounding in MLLMsLLM reasoning → Multimodal reasoning +1
Beyond Static Tools: Test-Time Tool Evolution for Scientific ReasoningLLM reasoning → Reasoning for agents, planning and code
Reinforcement Learning for Micro-Level Claims ReservingDeep reinforcement learning methods → Adversarial robustness and RL in finance
DIAGPaper: Diagnosing Valid and Specific Weaknesses in Scientific Papers via Multi-Agent ReasoningLLM reasoning → Knowledge- and tool-augmented reasoning +1
GRPO with State Mutations: Improving LLM-Based Hardware Test Plan GenerationLLM reasoning → Reasoning for agents, planning and code
Beyond Entangled Planning: Task-Decoupled Planning for Long-Horizon AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
d3LLM: Ultra-Fast Diffusion LLM using Pseudo-Trajectory DistillationEfficient LLM inference and serving → Diffusion LLM and MoE inference
VirtualEnv: A Platform for Embodied AI ResearchLLM reasoning → Spatial and embodied reasoning
From RAG to Agentic RAG for Faithful Islamic Question AnsweringLLM agents and tool use → Web, search, and research agents
MegaFlow: Large-Scale Distributed Orchestration System for the Agentic EraLLM agents and tool use → Software engineering coding agents
Thinking Before Constraining: A Unified Decoding Framework for Large Language ModelsLLM reasoning → Test-time compute and efficient thinking
Stagewise Reinforcement Learning and the Geometry of the Regret LandscapeDeep reinforcement learning methods → RL theory and sample complexity
Controlling Multimodal Conversational Agents with Coverage-Enhanced Latent ActionsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
Land-then-transport: A Flow Matching-Based Generative Decoder for Wireless Image TransmissionDiffusion models → Diffusion for compression and wireless
FROAV: A Framework for RAG Observation and Agent Verification -- Lowering the Barrier to LLM Agent ResearchLLM agents and tool use → Web, search, and research agents
Online Markov Decision Processes with Terminal Law ConstraintsDeep reinforcement learning methods → RL theory and sample complexity
FocalOrder: Focal Preference Optimization for Reading Order DetectionLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
JudgeFlow: Agentic Workflow Optimization via Block JudgeLLM agents and tool use → Software engineering coding agents
Learning How to Remember: A Meta-Cognitive Management Method for Structured and Transferable Agent MemoryLLM agents and tool use → Agent memory systems
Knowledge Distillation for LLM-Based Human Activity Recognition in HomesKnowledge distillation → Distillation for medical imaging
Beyond Dialogue Time: Temporal Semantic Memory for Personalized LLM AgentsLLM agents and tool use → Agent memory systems
IFDNS: An Iterative Feedback-Driven Neuro-Symbolic Method for Faithful Logical ReasoningLLM reasoning → Reasoning benchmarks and evaluation
Puzzle it Out: Local-to-Global World Model for Offline Multi-Agent Reinforcement LearningWorld models → Latent dynamics models and model-based RL +1
From Sketch to Fresco: Efficient Diffusion Transformer with Progressive ResolutionDiffusion models → Diffusion foundations and sampling
RLPO: Residual Listwise Preference Optimization for Long-Context Review RankingLLM alignment / RLHF / preference optimisation → Preference alignment for recommenders and RAG
SDHSI-Net: Learning Better Representations for Hyperspectral Images via Self-DistillationKnowledge distillation → Detection and dense prediction distillation
Outcome-Grounded Advantage Reshaping for Fine-Grained Credit Assignment in Mathematical ReasoningLLM reasoning → Reinforcement learning for reasoning
Forecast the Principal, Stabilize the Residual: Subspace-Aware Feature Caching for Efficient Diffusion TransformersDiffusion models → Diffusion foundations and sampling
MCP-ITP: An Automated Framework for Implicit Tool Poisoning in MCPLLM agents and tool use → Agent threats, guardrails and governance
On the Non-decoupling of Supervised Fine-tuning and Reinforcement Learning in Post-trainingDeep reinforcement learning methods → RL training of language and agent models
OpenTinker: Separating Concerns in Agentic Reinforcement LearningLLM agents and tool use → Reinforcement learning for LLM agents
GROKE: Vision-Free Navigation Instruction Evaluation via Graph Reasoning on OpenStreetMapLLM reasoning → Spatial and embodied reasoning
Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language ModelsMixture-of-experts → MoE large language models
Seeing Right but Saying Wrong: Inter- and Intra-Layer Refinement in MLLMs without TrainingMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
TALON: Confidence-Aware Speculative Decoding with Adaptive Token TreesEfficient LLM inference and serving → Speculative decoding
Beyond Hard Masks: Progressive Token Evolution for Diffusion Language ModelsDiffusion models → Discrete and text diffusion +1
Reward Modeling from Natural Language Human FeedbackLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Agentic Diagnostic Reasoning over Telecom and Datacenter InfrastructureLLM reasoning → Reasoning for agents, planning and code +1
BayesRAG: Probabilistic Mutual Evidence Corroboration for Multimodal Retrieval-Augmented GenerationRetrieval-augmented generation → Multimodal RAG
Segmental Advantage Estimation: Enhancing PPO for Long-Context LLM TrainingLLM reasoning → Reinforcement learning for reasoning
VLM-CAD: VLM-Optimized Collaborative Agent Design Workflow for Analog Circuit SizingLLM reasoning → Reasoning for agents, planning and code +2
ARM: Role-Conditioned Neuron Transplantation for Training-Free Generalist LLM Agent MergingLLM agents and tool use → General LLM agent frameworks and benchmarks
Heterogeneous Multi-Expert Reinforcement Learning for Long-Horizon Multi-Goal Tasks in Autonomous ForkliftsDeep reinforcement learning methods → RL for robot locomotion and manipulation
Mimic Human Cognition, Master Multi-Image Reasoning: A Meta-Action Framework for Enhanced Visual UnderstandingLLM reasoning → Multimodal reasoning +1
LRAS: Advanced Legal Reasoning with Agentic SearchLLM reasoning → Reasoning for agents, planning and code +1
Inference-Time Scaling for Visual AutoRegressive modeling by Searching Representative SamplesLLM reasoning → Diffusion reasoning and test-time scaling +1
A Visual Semantic Adaptive Watermark grounded by Prefix-Tuning for Large Vision-Language ModelMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
Focal Guidance: Unlocking Controllability from Semantic-Weak Layers in Video Diffusion ModelsDiffusion models → Image generation, text-to-image and editing
ReasonTabQA: A Comprehensive Benchmark for Table Question Answering from Real World Industrial ScenariosLLM reasoning → Knowledge- and tool-augmented reasoning

All 370