2026

dTTRL: Test-Time Reinforcement Learning for Diffusion Language Model

Mingyu Cao, Alvaro H.C. Correia, Christos Louizos, Handa Li, Shiwei Liu#, Lu Yin# (# corresponding author)

Submitted to Neurips 2026

A test-time reinforcement learning method for diffusion language models.

dTTRL: Test-Time Reinforcement Learning for Diffusion Language Model

Mingyu Cao, Alvaro H.C. Correia, Christos Louizos, Handa Li, Shiwei Liu#, Lu Yin# (# corresponding author)

Submitted to Neurips 2026

A test-time reinforcement learning method for diffusion language models.

Search or Accelerate: Confidence-Switched Position Beam Search for Diffusion Language Models

Mingyu Cao, Alvaro H.C. Correia, Christos Louizos, Shiwei Liu#, Lu Yin# (# corresponding author)

ICML 2026 Regular

SOAR is a training-free decoding algorithm for Diffusion Language Models that adaptively switches between wider search and parallel decoding based on model confidence, improving reasoning and code generation quality without sacrificing inference speed.

[Code] [Page] [Paper]

Search or Accelerate: Confidence-Switched Position Beam Search for Diffusion Language Models

Mingyu Cao, Alvaro H.C. Correia, Christos Louizos, Shiwei Liu#, Lu Yin# (# corresponding author)

ICML 2026 Regular

SOAR is a training-free decoding algorithm for Diffusion Language Models that adaptively switches between wider search and parallel decoding based on model confidence, improving reasoning and code generation quality without sacrificing inference speed.

[Code] [Page] [Paper]

Condense, Don't Just Prune: Enhancing Efficiency and Performance in MoE Layer Pruning

Mingyu Cao, Gen Li, Jie Ji, Jiaqi Zhang, Ajay Jaiswal, Li Shen, Xiaolong Ma, Shiwei Liu, Lu Yin# (# corresponding author)

TMLR 2026

A pruning method for Mixture-of-Experts models that merges multiple experts per layer into a reduced set, preserving model quality while reducing memory usage and improving inference efficiency.

[Code] [Page] [Paper]

Condense, Don't Just Prune: Enhancing Efficiency and Performance in MoE Layer Pruning

Mingyu Cao, Gen Li, Jie Ji, Jiaqi Zhang, Ajay Jaiswal, Li Shen, Xiaolong Ma, Shiwei Liu, Lu Yin# (# corresponding author)

TMLR 2026

A pruning method for Mixture-of-Experts models that merges multiple experts per layer into a reduced set, preserving model quality while reducing memory usage and improving inference efficiency.

[Code] [Page] [Paper]

2020

A neural network-based joint learning approach for biomedical entity and relation extraction from biomedical literature

Ling Luo, Zhihao Yang, Mingyu Cao, Lei Wang, Yin Zhang, Hongfei Lin

Journal of biomedical informatics (JBI) 2020-03-01

Joint extraction of entities and relations for biomedical text.

[Paper]

A neural network-based joint learning approach for biomedical entity and relation extraction from biomedical literature

Ling Luo, Zhihao Yang, Mingyu Cao, Lei Wang, Yin Zhang, Hongfei Lin

Journal of biomedical informatics (JBI) 2020-03-01

Joint extraction of entities and relations for biomedical text.

[Paper]