Papers by Junwei Sheng

3 papers
Compatibility-Aware Dynamic Fine-Tuning for Large Language Models (2026.acl-long)

Copied to clipboard

Challenge: Recent work attributes optimization instability to the low probability of demonstrations being incompatible with the sample level.
Approach: They propose a Dynamic Fine-Tuning extension of DFT that controls sample-level optimization variance.
Outcome: The proposed model can generalize token-level stabilization to the sample level while remaining fully supervised and free of reward modeling.
Pathway2Text: Dataset and Method for Biomedical Pathway Description Generation (2022.findings-naacl)

Copied to clipboard

Challenge: Neural text generation is a novel technique to describe biomedical pathways without manually curation.
Approach: They propose a new dataset Pathway2Text which contains 2,367 pairs of biomedical pathways and textual descriptions.
Outcome: The proposed method improves on both Graph2Text and Text2Graph tasks and can be used as a benchmark for biomedical named entity recognition.
MetaFill: Text Infilling for Meta-Path Generation on Heterogeneous Information Networks (2022.emnlp-main)

Copied to clipboard

Challenge: Existing meta-path generation methods cannot fully exploit rich textual information in HINs.
Approach: They propose a text-infilling-based approach to generate meta-paths from textual information in HINs.
Outcome: The proposed approach can classify edges in the zero-shot setting, where existing methods cannot generate meta-paths.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations