Papers by Desheng Wu
Leveraging Human and Machine Preferences for Zero-shot Detection of AI-Generated Text (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent advances in large language models have enabled generated texts to closely mimic human writing, posing significant challenges to the detection of AI-generated content. |
| Approach: | They propose a human-machine prediction discrepancy adapter for AI-generated text detection . they use a joint fine-tuning strategy and a discrepany-aware reweighting mechanism . |
| Outcome: | The proposed framework improves the detection performance of five representative models under various evaluation scenarios. |
Extending First-Order Logic for Factual Reasoning over Knowledge Graphs (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for factual reasoning over knowledge graphs lack support for multiple quantifiers and connectives. |
| Approach: | They propose an extended FOL -structure over knowledge graphs that incorporates comparison predicates and counting quantifiers. |
| Outcome: | The proposed method achieves state-of-the-art on Fact-FOLX-KG, while previous methods experience performance drop on claims requiring comparison and counting. |
SenDetEX: Sentence-Level AI-Generated Text Detection for Human-AI Hybrid Content via Style and Context Fusion (2025.emnlp-main)
Copied to clipboard
| Challenge: | Text generated by Large Language Models (LLMs) now rivals human writing, raising concerns about its misuse. |
| Approach: | They propose a framework for sentence-level AI-generated text detection via style and context fusion. |
| Outcome: | The proposed framework outperforms baseline models in detection accuracy while exhibiting transferability and robustness. |
Fact Verification on Knowledge Graph via Programmatic Graph Reasoning (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for fact verification on knowledge graphs use implicit reasoning to predict entailment between claims and KG triples. |
| Approach: | They propose a framework that integrates large language models for fact verification on knowledge graphs. |
| Outcome: | The proposed framework outperforms existing methods on knowledge graphs with 86.82% accuracy. |