Jun Wang

Bottlenecked Transformers: Periodic KV Cache Consolidation for Generalised Reasoning

ViMo: A Generative Visual GUI World Model for App Agents

SpatialViz-Bench: A Cognitively-Grounded Benchmark for Diagnosing Spatial Visualization in MLLMs

CurES: From Gradient Analysis to Efficient Curriculum Learning for Reasoning LLMs

AccidentBench: Benchmarking Multimodal Understanding and Reasoning in Vehicle Accidents and Beyond

VoG: Enhancing LLM Reasoning through Stepwise Verification on Knowledge Graphs

A Benchmark for Deep Information Synthesis

From Experience to Strategy: Empowering LLM Agents with Trainable Graph Memory

SuRe: Surprise-Driven Prioritised Replay for Continual LLM Learning

Paying Attention to Hybrid Attention: Untangling the Issues with Conversion Methods

Grouped-head latenT Attention

LoopServe: An Adaptive Dual-phase LLM Inference Acceleration System for Multi-Turn Dialogues

Hi-Agent: Hierarchical Vision-Language Agents for Mobile Device Control

Memory-Driven Self-Improvement with Large Language Models

Evolving LLMs' Self-Refinement Capability via Synergistic Training-Inference Optimization

Emergent Bayesian Behaviour and Optimal Cue Combination in LLMs

BEYOND SYNTAX: ACTION SEMANTICS LEARNING FOR APP AGENTS

Activation-aware Probe-Query: Effective Key-Value Retrieval for Long-Context LLMs Inference

Advancing Autonomous VLM Agents via Variational Subgoal-Conditioned Reinforcement Learning

Subjective Depth and Timescale Transformers: Learning Where and When to Compute

Distill Not Only Data but Also Rewards: Can Smaller Language Models Surpass Larger Ones?

Causal Discovery under Changing Mechanisms: A Unified Graphical Approach

MF-LLM: Simulating Population Decision Dynamics via a Mean-Field Large Language Model Framework

Lightweight Neural App Control

SPA-BENCH: A COMPREHENSIVE BENCHMARK FOR SMARTPHONE AGENT EVALUATION

Succeed or Learn Slowly: Sample Efficient Off-Policy Reinforcement Learning for Mobile App Control

Mixture of Attentions For Speculative Decoding

Risk-aware Direct Preference Optimization under Nested Risk Measure

GenPO: Generative Diffusion Models Meet On-Policy Reinforcement Learning

Causal Sufficiency and Necessity Improves Chain-of-Thought Reasoning

Uncertainty-quantified Rollout Policy Adaptation for Unlabelled Cross-domain Video Temporal Grounding

Self-Verifying Reflection Helps Transformers with CoT Reasoning

DistRL: An Asynchronous Distributed Reinforcement Learning Framework for On-Device Control Agent

Circuit Transformer: A Transformer That Preserves Logical Equivalence

A Principle of Targeted Intervention for Multi-Agent Reinforcement Learning

Curious Causality-Seeking Agents Learn Meta Causal World

ReMA: Learning to Meta-Think for LLMs with Multi-agent Reinforcement Learning

On the Optimization Landscape of Low Rank Adaptation Methods for Large Language Models

Efficient Reinforcement Learning with Large Language Model Priors

Human-inspired Episodic Memory for Infinite Context LLMs

Direct Preference Optimization Using Sparse Feature-level Constraints

Attaining Human's Desirable Outcomes in Indirect Human-AI Interaction via Multi-Agent Influence Diagrams

Constrain Alignment with Sparse Autoencoders

Large Language Models are Demonstration Pre-Selectors for Themselves

SparsePO: Controlling Preference Alignment of LLMs via Sparse Token Masks

Risk-aware Direct Preference Optimization under Nested Risk Measure

Large Language Models are Demonstration Pre-Selectors for Themselves

Mitigating Unobserved Confounding via Diffusion Probabilistic Models

Active Causal Learning for Conditional Average Treatment Effect Estimation

Temporal Visiting-Monitoring Feature Interaction Learning for Modelling Structured Electronic Health Records

FedADM: Adaptive Federated Learning via Dissimilarity Measure