Publications

Global Credit Assignment via Dynamical Criticality

First author | Peking University

ICML 2026 | Accepted

This ICML 2026 paper studies temporal credit assignment in recurrent, convolutional recurrent, and spiking systems through criticality-driven online local learning with constant activation memory.

Research & contributions

• Studies temporal credit assignment beyond BPTT through a criticality-driven local learning rule.
• Uses long-range spatial inertia and temporal correlations to approximate global credit propagation with a locally computable teaching signal.
• Derives a closed-form update requiring constant activation memory and O(H) auxiliary state, and validates accuracy, stability, and scalability on RNN, ConvRNN, and SNN benchmarks.

Decision-Making Large Language Model for Wireless Communication: A Comprehensive Survey on Key Techniques

Ning Yang, Mingrui Fan, Wentao Wang, Haijun Zhang

Third author (advisor is first author) | Institute of Automation, Chinese Academy of Sciences

IEEE COMST | Published

This survey organizes decision-making LLMs for wireless communication across data construction, architecture adaptation, reasoning control, multi-agent coordination, and open challenges.

Research & contributions

• Surveys LLM-enabled decision-making for wireless communication across data, modeling, reasoning, inference, learning-based control, and multi-agent settings.
• I independently contributed the sections on data generation and augmentation, architecture adaptation, and open challenges.
• The paper was published in IEEE Communications Surveys & Tutorials.

Cooperative Edge Caching with Large Language Model in Wireless Networks

Ning Yang, Wentao Wang, Lingtao Ouyang, Haijun Zhang

Second author (student first; advisor is first author) | Institute of Automation, Chinese Academy of Sciences

IEEE TMC | Accepted

We formulate cooperative multi-base-station edge caching as an LLM-native sequential decision problem, build an SFT+GRPO training pipeline, and design an opportunity-aware reward.

Research & contributions

• Builds a multi-BS, multi-user cooperative edge-caching environment and formulates it as an LLM-native sequential decision problem.
• Implements a two-stage SFT+GRPO training pipeline with TRL, Unsloth, and QLoRA under a strictly validated text-to-action interface.
• Achieves strong generalization across user scale, base-station count, and request distributions.

MC-TRCM: Observation-Aware Recursive Fusion for Incomplete Mobile and Wearable Mental-Health Feature Views

First author | Dalian University of Technology

ICONIP 2026 | Accepted

Observation-aware recursive fusion for asynchronous and incomplete mobile and wearable feature views, combining shared-private representations with task-conditioned multi-task prediction.

Research & contributions

• Encodes each feature source independently and explicitly models missingness.
• Combines shared-private latent representations, task-conditioned fusion, and recursive prediction.
• Achieves leading or competitive results on six tasks across DepreST-CAT and PSYCHE-D.

LinearARD: Linear-Memory Attention Distillation for RoPE Restoration

Co-first author (advisor listed first) | Institute of Automation, Chinese Academy of Sciences

NeurIPS 2026 | Accepted

LinearARD aligns row-wise dense self-relations between native-RoPE teachers and RoPE-scaled students, restoring long-context performance with an exact linear-memory distillation objective.

Research & contributions

• Proposes a restoration-distillation framework that aligns row-wise dense self-relations between a native-RoPE teacher and a RoPE-scaled student.
• Derives an exact linear-memory KL kernel, avoiding explicit construction of full relation matrices on long sequences.
• On LLaMA2-7B scaled from 4K to 32K, the method recovers 98.3% of short-context performance using only 4.25M training tokens.
• Responsible for writing the full manuscript.

Cluster-Aware Attention-Based Deep Reinforcement Learning for Pickup and Delivery Problems

Wentao Wang, Lifeng Han, Guangyu Zou

First author | Dalian University of Technology

Applied Intelligence | Major Revision

CAADRL combines global self-attention, intra-cluster attention, and a dynamic dual-decoder to exploit clustered structure in pickup and delivery problems while reducing inference latency.

Research & contributions

• Proposes CAADRL, a cluster-aware deep reinforcement learning framework for pickup and delivery problems with clustered structure.
• Integrates global self-attention, intra-cluster attention, and a dynamic dual-decoder with a learnable gate.
• Achieves leading performance on clustered benchmarks while reducing inference latency relative to neural collaborative-search baselines.