Invisible Tokens, Visible Bills: The Urgent Need to Audit Hidden Operations in Opaque LLM Services
Published in ICML 2026, Under Review, 2026
Invisible Tokens, Visible Bills: The Urgent Need to Audit Hidden Operations in Opaque LLM Services.
Publications
Selected papers are shown below. Refer to Google Scholar for more details.
CO-EVOLVE: Bidirectional Co-Evolution of Graph Structure and Semantics for Heterophilous Learning
Published in Working Paper, 2026
Existing LLM-GNN integration relies on static, unidirectional pipelines, causing bidirectional error propagation. We introduce a Dual-View Co-Evolution mechanism where the GNN injects topological context to guide the LLM, and the LLM constructs a dynamic semantic graph.
A Reinforcement Learning Framework for Application-Specific TCP Congestion-Control
Published in IPCCC 2025, 2025
This paper proposes a DRL based TCP congestion control framework, which caters to handle various objectives (maximize throughput, minimize latency, …) set by various applications.
NetSight: Graph Attention Based Network Traffic Forecasting
Published in ICDCS 2026, Under Review, 2025
The paper introduces NetSight, a model that captures global-local spatial-temporal dependencies through a data-driven fusion graph, node normalization, and multi-head attention, significantly improving network traffic forecasting accuracy.
Weighted Fuzzy Rough Sets-based Tri-training and Its Application to Medical Diagnosis
Published in Applied Soft Computing, 2022
In this paper, for semi-supervised learning, we propose a new framework WFRSTT, which incorporates the merits of Fuzzy Rough Sets, Tri-training, a novel noise-filtering technique bad-points, and the multi-view information. Experiment results demonstrate its effectiveness in Medical Diagnosis.