Institution profile

Ocean University of China

Academic institutionasia · cn
Official website
Research library194linked papers
Opportunities0open roles
Selected work

Representative Papers

AppellateGen: A Benchmark for Appellate Legal Judgment Generation

Jan 04, 2026arXiv.org

Existing research on legal judgment generation has predominantly focused on first-instance trials, overlooking the complex dialectical reasoning required in appellate proceedings that involve evaluating initial judgments alongside new evidence. This work addresses this gap by introducing AppellateGen, the first benchmark dataset for appellate judgment generation, comprising 7,351 paired case records. We further propose a Standard Operating Procedure (SOP)-driven Legal Multi-Agent System (SLMAS) that decomposes judgment generation into distinct stages—issue identification, evidence retrieval, and opinion drafting—explicitly modeling the causal dependencies among judicial phases. Experimental results demonstrate that SLMAS significantly enhances the logical consistency of generated judgments, though large language models still face notable challenges in handling the intricate reasoning demands of appellate adjudication. The dataset and code are publicly released.

1 citationsRead paper

Forensics Adapter: Adapting CLIP for Generalizable Face Forgery Detection

Nov 29, 2024arXiv.org

Existing CLIP-based face forgery detection methods treat CLIP solely as a static feature extractor, lacking task-specific adaptation capability and suffering from limited generalization. To address this, we propose Forensics Adapter—a lightweight, forgery-boundary-aware adapter (5.7M parameters) that guides CLIP to explicitly model and fuse forensic traces. We further introduce a cross-module visual token interaction mechanism to enhance inter-layer propagation of localized forgery cues. Our framework marks the first paradigm shift transforming CLIP from a generic feature extractor into a task-driven, adaptable forgery detector. Evaluated on five standard benchmarks, our method achieves an average accuracy improvement of approximately 7% over state-of-the-art CLIP-based baselines, establishing a new benchmark for CLIP-based face forgery detection.

1 citationsRead paper
Recent publications

Latest Papers