Ruimin Hu

Decheng Liu, Zongqi Wang, Chunlei Peng et al.

Due to the successful development of deep image generation technology, forgery detection plays a more important role in social and economic security. Racial bias has not been explored thoroughly in the deep forgery detection field. In the paper, we first contribute a dedicated dataset called the Fair Forgery Detection (FairFD) dataset, where we prove the racial bias of public state-of-the-art (SOTA) methods. Different from existing forgery detection datasets, the self-constructed FairFD dataset contains a balanced racial ratio and diverse forgery generation images with the largest-scale subjects. Additionally, we identify the problems with naive fairness metrics when benchmarking forgery detection models. To comprehensively evaluate fairness, we design novel metrics including Approach Averaged Metric and Utility Regularized Metric, which can avoid deceptive results. We also present an effective and robust post-processing technique, Bias Pruning with Fair Activations (BPFA), which improves fairness without requiring retraining or weight updates. Extensive experiments conducted with 12 representative forgery detection models demonstrate the value of the proposed dataset and the reasonability of the designed fairness metrics. By applying the BPFA to the existing fairest detector, we achieve a new SOTA. Furthermore, we conduct more in-depth analyses to offer more insights to inspire researchers in the community.

Ruimin Hu, Mehmet Mercangöz

Manufacturing systems exhibit strong concurrency, synchronization, and contention for shared reusable resources, which makes fast and reliable scheduling and verification challenging. Petri nets provide a rigorous formalism for modeling such discrete-event manufacturing systems, but reachability analysis and solving remain difficult for conventional graph search or optimization-based solvers, particularly under state-space explosion and evolving production requirements. Recently, Large language models (LLMs) have shown promise as flexible planners that can generate candidate action sequences from textual specifications. However, direct use of LLMs for Petri net reachability remains unreliable. This paper proposes an LLM-based solving framework augmented with a two-stage reflection and reprompting mechanism. The combined effects of reflection and re-clarification improve the accuracy of feasible sequence generation. The proposed method is evaluated on an industrial case modeled as a Petri net. Under a fixed Petri net structure, the proposed strategy is assessed on six solvable reachability configurations. The results demonstrate improved reliability and stability in solving Petri net reachability problems. The proposed framework is further evaluated across multiple LLMs, which indicates that the framework is not tied to any specific model.