Papers

New on arXiv

Fri 10 Apr 2026

432 papers

GenesisFunc: Multi-Agent Data Generation for Accurate and Generalizable Function-CallingLLM agents and tool use → Tool learning and function calling
Aryabhata 2: Scaling Reinforcement Learning for Advanced STEM ReasoningLLM reasoning → Reinforcement learning for reasoning
Micro-Macro Retrieval: Reducing Long-Form Hallucination in Large Language ModelsLLM reasoning → Reasoning benchmarks and evaluation
Personalized Observation Normalization for Federated Reinforcement Learning in Simulation Environments with HeterogeneityDeep reinforcement learning methods → RL for energy, networks and operations
Bridging the Stability-Expressivity Gap: Synthetic Data Scaling and Preference Alignment for Low-Resource Spoken Language ModelsLLM alignment / RLHF / preference optimisation → Preference alignment for recommenders and RAG +1
The Alignment Floor: How Persona Customization Breaks Safety in Weakly-Aligned LLMsLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
From Intent to AI Pipelines: A Controlled Agentic Framework for Non-AI Expert ScientistsLLM agents and tool use → Web, search, and research agents
Query-Conditioned Graph Retrieval for Contextualized LLM Reasoning in Personalized Wearable DataLLM reasoning → Clinical and biomedical applications
ALDEN: Boosting Private Data Extraction from Retrieval-Augmented Generation Systems via Active Learning and Distribution EstimationRetrieval-augmented generation → Privacy and security of RAG
Improving Medical VQA through Trajectory-Aware Process SupervisionLLM reasoning → Clinical and biomedical applications
Single-Position Intervention Fails: Distributed Output Templates Drive In-Context LearningMechanistic interpretability → Circuits and mechanistic analysis
Continual Distillation of Teachers from Different DomainsKnowledge distillation → General knowledge distillation methods
MP-ISMoE: Mixed-Precision Interactive Side Mixture-of-Experts for Efficient Transfer LearningMixture-of-experts → MoE large language models
DeepTutor: Towards Agentic Personalized TutoringLLM agents and tool use → Educational agents and math problem solving
Sequential KV Cache Compression via Probabilistic Language Tries: Beyond the Per-Vector Shannon LimitEfficient LLM inference and serving → KV cache optimization
PriHA: A RAG-Enhanced LLM Framework for Primary Healthcare Assistant in Hong KongRetrieval-augmented generation → Medical and biomedical RAG
Can Coding Agents Be General Agents?LLM agents and tool use → Software engineering coding agents
Fairness in Multi-Agent Systems for Software Engineering: An SDLC-Oriented Rapid ReviewLLM agents and tool use → Software engineering coding agents
M^star: Every Task Deserves Its Own Memory HarnessLLM agents and tool use → Agent memory systems
Unmixing-Guided Spatial-Spectral Mamba with Clustering Tokens for Hyperspectral Image ClassificationState-space sequence models (S4, Mamba) → Hyperspectral image analysis with Mamba
HealthAdminBench: Evaluating Computer-Use Agents on Healthcare Administration TasksLLM agents and tool use → Medical and scientific domain agents
A Tale of Two Temperatures: Simple, Efficient, and Diverse Sampling from Diffusion Language ModelsDiffusion models → Discrete and text diffusion
Toward Explanatory Equilibrium: Verifiable Reasoning as a Coordination Mechanism under Asymmetric InformationLLM reasoning → Reasoning for agents, planning and code +1
From UAV Imagery to Agronomic Reasoning: A Multimodal LLM Benchmark for Plant PhenotypingLLM reasoning → Clinical and biomedical applications +1
Should We be Pedantic About Reasoning Errors in Machine Translation?LLM reasoning → Reasoning benchmarks and evaluation
Genie 4D: Semantic-Prior-Guided 4D Dynamic Scene ReconstructionWorld models → Embodied and robot world models
Relational Preference Encoding in Looped Transformer Internal StatesLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Evolutionary Token-Level Prompt Optimization for Diffusion ModelsDiffusion models → Image generation, text-to-image and editing
MEMENTO: Teaching LLMs to Manage Their Own ContextLLM reasoning → Test-time compute and efficient thinking +1
Training-Free Object-Background Compositional T2I via Dynamic Spatial Guidance and Multi-Path PruningDiffusion models → Image generation, text-to-image and editing
Is There Knowledge Left to Extract? Evidence of Fragility in Medically Fine-Tuned Vision-Language ModelsMultimodal LLMs / vision-language models → Medical vision-language models
Steered LLM Activations are Non-SurjectiveMechanistic interpretability → Activation steering methods
Vector Field Synthesis with Sparse Streamlines Using Diffusion ModelDiffusion models → Inverse problems and image restoration
ProGAL-VLA: Grounded Alignment through Prospective Reasoning in Vision-Language-Action ModelsVision-language-action models → Generalist VLA models and benchmarks
EE-MCP: Self-Evolving MCP-GUI Agents via Automated Environment Generation and Experience LearningLLM agents and tool use → GUI and mobile agents +1
Controllable and Verifiable Tool-Use Data Synthesis for Agentic Reinforcement LearningLLM agents and tool use → Reinforcement learning for LLM agents +1
Building an Internal Coding Agent at Zup: Lessons and Open QuestionsLLM agents and tool use → Software engineering coding agents
Pioneer Agent: Continual Improvement of Small Language Models in ProductionLLM agents and tool use → General LLM agent frameworks and benchmarks
Text-Guided 6D Object Pose Rearrangement via Closed-Loop VLM AgentsLLM reasoning → Spatial and embodied reasoning +1
The Myth of Expert Specialization in MoEs: Why Routing Reflects Geometry, Not Necessarily Domain ExpertiseMixture-of-experts → Expert routing strategies
MedLVR: Latent Visual Reasoning for Reliable Medical Visual Question AnsweringLLM reasoning → Clinical and biomedical applications
Conflicts Make Large Reasoning Models Vulnerable to AttacksLLM reasoning → Reasoning for agents, planning and code
See Fair, Speak Truth: Equitable Attention Improves Grounding and Reduces Hallucination in Vision-Language AlignmentMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
Backdoors in RLVR: Jailbreak Backdoors in LLMs From Verifiable RewardLLM reasoning → Reinforcement learning for reasoning +1
ADAM: A Systematic Data Extraction Attack on Agent Memory via Adaptive QueryingLLM agents and tool use → Agent threats, guardrails and governance +1
CONSCIENTIA: Can LLM Agents Learn to Strategize? Emergent Deception and Trust in a Multi-Agent NYC SimulationLLM agents and tool use → Social simulation and multi-agent debate
MPAC: A Multi-Principal Agent Coordination Protocol for Interoperable Multi-Agent CollaborationLLM agents and tool use → General LLM agent frameworks and benchmarks
Tango: Taming Visual Signals for Efficient Video Large Language ModelsEfficient LLM inference and serving → Visual token pruning in multimodal models
Large Language Models Generate Harmful Responses Using a Distinct Mechanism, Shared Across Harm TypesLLM safety and security → Safety alignment, refusal and guardrails
Trans-RAG: Query-Centric Vector Transformation for Secure Cross-Organizational RetrievalRetrieval-augmented generation → Privacy and security of RAG
EgoTL: Egocentric Think-Aloud Chains for Long-Horizon TasksWorld models → World models for language agents
Seeing is Believing: Robust Vision-Guided Cross-Modal Prompt Learning under Label NoiseMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
VisionFoundry: Teaching VLMs Visual Perception with Synthetic ImagesMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
VL-Calibration: Decoupled Confidence Calibration for Large Vision-Language Models ReasoningLLM reasoning → Multimodal reasoning +1
Toward World Models for EpidemiologyWorld models → World models for language agents
RIRF: Reasoning Image Restoration FrameworkLLM reasoning → Multimodal reasoning
VISOR: Agentic Visual Retrieval-Augmented Generation via Iterative Search and Over-horizon ReasoningLLM reasoning → Multimodal reasoning +3
Physics-Informed Reinforcement Learning of Spatial Density Velocity Potentials for Map-Free RacingDeep reinforcement learning methods → RL for autonomous driving and traffic
RecaLLM: Addressing the Lost-in-Thought Phenomenon with Explicit In-Context RetrievalLLM reasoning → Knowledge- and tool-augmented reasoning
Process Reward Agents for Steering Knowledge-Intensive ReasoningRetrieval-augmented generation → Agentic RAG, agents and memory +1
Agentic Jackal: Live Execution and Semantic Value Grounding for Text-to-JQLLLM agents and tool use → Web, search, and research agents
From Reasoning to Agentic: Credit Assignment in Reinforcement Learning for Large Language ModelsLLM reasoning → Reinforcement learning for reasoning +1
E3-TIR: Enhanced Experience Exploitation for Tool-Integrated ReasoningLLM reasoning → Reasoning for agents, planning and code +1
SafeAdapt: Provably Safe Policy Updates in Deep Reinforcement LearningDeep reinforcement learning methods → Safe and constrained RL
Many-Tier Instruction Hierarchy in LLM AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
UIPress: Bringing Optical Token Compression to UI-to-Code GenerationMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
SCoRe: Clean Image Generation from Diffusion Models Trained on Noisy ImagesDiffusion models → Diffusion foundations and sampling
On the Representational Limits of Quantum-Inspired 1024-D Document Embeddings: An Experimental Evaluation FrameworkRetrieval-augmented generation → Generation and context integration
Do Vision Language Models Need to Process Image Tokens?Multimodal LLMs / vision-language models → Visual token pruning and efficiency
Sharp description of local minima in the loss landscape of high-dimensional two-layer ReLU neural networksKnowledge distillation → General knowledge distillation methods
Do AI Coding Agents Log Like Humans? An Empirical StudyLLM agents and tool use → Software engineering coding agents
HiL-Bench (Human-in-Loop Benchmark): Do Agents Know When to Ask for Help?LLM agents and tool use → General LLM agent frameworks and benchmarks
EGLOCE: Training-Free Energy-Guided Latent Optimization for Concept ErasureDiffusion models → Robustness, privacy and adversarial aspects
Efficient Unlearning through Maximizing Relearning Convergence DelayLLM safety and security → Machine unlearning
BadSkill: Backdoor Attacks on Agent Skills via Model-in-Skill PoisoningLLM safety and security → Prompt injection and agent security
Task-Aware LLM Routing with Multi-Level Task-Profile-Guided Data Synthesis for Cold-Start ScenariosSynthetic data and self-improvement for LLMs → Code and instruction data synthesis
EpiAgent: An Agent-Centric System for Ancient Inscription RestorationLLM agents and tool use → Multimodal and embodied agents
Arbitration Failure, Not Perceptual Blindness: How Vision-Language Models Resolve Visual-Linguistic ConflictsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning +1
Visually-Guided Policy Optimization for Multimodal ReasoningLLM reasoning → Multimodal reasoning +1
Mind the Gap Between Spatial Reasoning and Acting! Step-by-Step Evaluation of Agents With Spatial-GymLLM reasoning → Spatial and embodied reasoning
VAG: Dual-Stream Video-Action Generation for Embodied Data SynthesisWorld models → Embodied and robot world models +1
Compositional-Degradation UAV Image Restoration: Conditional Decoupled MoE Network and A BenchmarkMixture-of-experts → Vision mixture-of-experts
Constraint-Aware Corrective Memory for Language-Based Drug Discovery AgentsLLM agents and tool use → Medical and scientific domain agents
SkillMOO: Multi-Objective Optimization of Agent Skills for Software EngineeringLLM agents and tool use → Software engineering coding agents
SAGE: A Service Agent Graph-guided Evaluation BenchmarkLLM agents and tool use → General LLM agent frameworks and benchmarks
Mosaic: Multimodal Jailbreak against Closed-Source VLMs via Multi-View Ensemble OptimizationMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
DRBENCHER: Can Your Agent Identify the Entity, Retrieve Its Properties and Do the Math?LLM agents and tool use → Web, search, and research agents
2D or 3D: Who Governs Salience in VLA Models? -- Tri-Stage Token Pruning Framework with Modality Salience AwarenessVision-language-action models → VLA architectures and generalization +1
Unreal Thinking: Chain-of-Thought Hijacking via Two-stage BackdoorLLM safety and security → Backdoor and data poisoning attacks
Training-free, Perceptually Consistent Low-Resolution Previews with High-Resolution Image for Efficient Workflows of Diffusion ModelsDiffusion models → Flow matching, time series and science
GRM: Utility-Aware Jailbreak Attacks on Audio LLMs via Gradient-Ratio MaskingLLM safety and security → Audio jailbreaks and speech backdoors
SHIFT: Steering Hidden Intermediates in Flow TransformersDiffusion models → Image generation, text-to-image and editing +1
SPASM: Stable Persona-driven Agent Simulation for Multi-turn Dialogue GenerationLLM agents and tool use → Social simulation and multi-agent debate
On the Role of DAG topology in Energy-Aware Cloud Scheduling : A GNN-Based Deep Reinforcement Learning ApproachDeep reinforcement learning methods → RL for energy, networks and operations
CT-1: Vision-Language-Camera Models Transfer Spatial Reasoning Knowledge to Camera-Controllable Video GenerationLLM reasoning → Multimodal reasoning +1
Do LLMs Follow Their Own Rules? A Reflexive Audit of Self-Stated Safety PoliciesLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks
LatentFlowSR: High-Fidelity Audio Super-Resolution via Noise-Robust Latent Flow MatchingDiffusion models → Audio, speech and music
MixFlow: Mixed Source Distributions Improve Rectified FlowsDiffusion models → Flow matching, time series and science
Generalization and Scaling Laws for Mixture-of-Experts TransformersMixture-of-experts → Expert routing strategies
Facet-Level Tracing of Evidence Uncertainty and Hallucination in RAGRetrieval-augmented generation → RAG robustness and adaptive retrieval

All 432