Wei-Lun Zhao

dblp:151/6208 · also Weilun Zhao · DBLP profile ↗
← Back
5ranked-venue papers
1as first author
2since 2021 · last 2026
—ORCID · none

Domains — the database's venue-derived domains; a paper can count in several

Artificial intelligence and machine learning · 2 · 2 since 2021

Expertise — from the expertise taxonomy: the topics of the expert's papers under the CCF categories. A weight counts papers with recency: 1 for a paper about the topic, 0.3 when the topic is its context, halved every five years.

Artificial intelligence
2 papers
Efficient and distributed learning · 72% Information extraction and text analysis · 28%
Software engineering, system software, and programming languages
1 paper
Empirical software engineering · 100%

Topics — the 5 heaviest of 5, each with the papers that count most for it

TopicWeightPapersLastEvidence papers
Natural language and speech › Information extraction and text analysis › document understanding
scientific document understanding
1.012026
AutoReproduce: Automatic AI Experiment Reproduction with Paper Lineage · ACL (1) 2026
Empirical software engineering
reproducibility
1.012026
AutoReproduce: Automatic AI Experiment Reproduction with Paper Lineage · ACL (1) 2026
Machine learning › Efficient and distributed learning
inference acceleration
0.912025
FR-Spec: Accelerating Large-Vocabulary Language Models via Frequency-Ranked Speculative Sampling · ACL (1) 2025
Machine learning › Efficient and distributed learning
inference efficiency
0.912025
FR-Spec: Accelerating Large-Vocabulary Language Models via Frequency-Ranked Speculative Sampling · ACL (1) 2025
Machine learning › Efficient and distributed learning › inference acceleration
speculative decoding
0.912025
FR-Spec: Accelerating Large-Vocabulary Language Models via Frequency-Ranked Speculative Sampling · ACL (1) 2025

Methods — techniques the papers use, named apart from their topics

paper lineage analysis · 2.0frequency-ranked speculative sampling · 0.9
YearPublicationVenuePosition
2026 AutoReproduce: Automatic AI Experiment Reproduction with Paper Lineage
abstract
Xuanle Zhao, Zilin Sang, Yuxuan Li, Qi Shi, Weilun Zhao, Shuo Wang, Duzhen Zhang, Xu Han, Zhiyuan Liu, Maosong Sun. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.
Xuanle Zhao, Zilin Sang, Qi Shi 0002, Wei-Lun Zhao, Shuo Wang 0013, Duzhen Zhang, Xu Han 0007, Zhiyuan Liu 0001, Maosong Sun 0001
ACL (1)5
2025 FR-Spec: Accelerating Large-Vocabulary Language Models via Frequency-Ranked Speculative Sampling
abstract
Weilin Zhao, Tengyu Pan, Xu Han, Yudi Zhang, Sun Ao, Yuxiang Huang, Kaihuo Zhang, Weilun Zhao, Yuxuan Li, Jie Zhou, Hao Zhou, Jianyong Wang, Maosong Sun, Zhiyuan Liu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.
Weilin Zhao, Tengyu Pan, Xu Han 0007, Sun Ao, Yuxiang Huang 0001, Kaihuo Zhang, Wei-Lun Zhao, Jie Zhou 0016, Hao Zhou 0012, Jianyong Wang 0001, Maosong Sun 0001, Zhiyuan Liu 0001
ACL (1)8
2014 A plasmonic refractive index sensor based on a MIM waveguide with a side-coupled nanodisk resonator
abstract
Based on a metal-insulator-metal (MIM) waveguide with a side-coupled nanodisk cavity, the sensor using the surface plasmon polaritons (SPPs) refractive index is investigated and studied numerically. The finite-difference time-domain (FDTD) method is used to simulate the performance of the sensor. The numerical simulation result indicates that all the resonance wavelengths in the transmission characteristic of the structure have a linear relationship with the refractive index of the cavity. Furthermore, the sensitivities of the sensor in this paper for the refractive index can be achieved as high as 1320 nm RIU for the mode1, 812.5 nm RIU for the mode2, 600 nm RIU for the mode3, respectively. Besides, the influences of the structural parameters on the transmission characteristic and the sensing characteristic are also studied in detail by the FDTD method. The sensor with compact and simple structure not only can be used to measure the temperature based on the linear relation between the refractive index and temperature, but also has many potential applications in optical networks on chip and On-chip sensor networks.
Ye-Xiong Huang, Yiyuan Xie, Wei-Lun Zhao, Hong-Jun Che, Jia-Chao Li
RTCSA3
2014 Performance optimization in Torus-based optical networks-on-chip
abstract
In this paper, the insertion loss and crosstalk noise of M × N Torus-based optical networks-on-chip (ONoCs) is systematically analyzed, which caused performance degradation. The proposed analysis model can be applied to arbitrary 5×5 routers and Torus-based ONoCs. When traditional non-blocking five-port optical routers used in the original Torus structure, it's suffered lager Bit Error Rate (BER) in a small scale. The router optimization and angle optimization method is used for achieving a better quality of network communication and performance improvement. The numerical results show the signal-to-noise ratio (SNR) of the worst-case transmission link in Torus-based ONoCs with certain size. When the network scale of Torus-based ONoC is 6×6 and the input power is 0 dB, the SNR of Torus-based ONoC using Crux router is 21.66 dB, which is 7.06 dB higher than optimized Crossbar router. With angle optimization further used in router level and network level, the SNR can reach to 23.87 dB. Moreover, we also find that a better SNR can be got with M gradually close to N .
Yiyuan Xie, Hong-Jun Che, Wei-Lun Zhao, Ye-Xiong Huang, Jia-Chao Li
RTCSA4
2014 Performance improvement in mesh-based optical networks-on-chip
abstract
In the optical communication system, crosstalk noise is always the critical factor affecting the optical signal transmission, especially in networks-on-chip (ONoCs). Based on the model at device, router and network level, this paper proposes the Optimized Crux (OC) router which uses the crossing angle of 60° or 120° instead of the conventional 90° to optimize the Crux optical router. The SNR of the Optimized Crux (OC) router is improved by 2.1dB compared with the Crux on the premise that the size of mesh-based ONoCs is 7×7. By comparing analysis, the results show that to achieve the bit error rate (BER) of 10−9for reliable transmissions, the maximum mesh-based ONoCs size has expanded from 4×4 when using the Crossbar optical router to 8×8 when using the OC router.
Wei-Lun Zhao, Yiyuan Xie, Hong-Jun Che, Ye-Xiong Huang, Jia-Chao Li
RTCSA1