graph-based-methods

该论文提出局部预测一致性（local predictive consistency）作为成对个体公平性定义的节点级松弛，通过相似度加权的邻域平均预测来生成可解释的公平性分数，并将其扩展到二分类、回归和生存分析场景。

TIGPO proposes a temporal instance-graph policy optimization method that extends graph-based credit assignment across policy updates for long-horizon LLM agents, using persistent transition graphs and revisit slots to improve advantage estimation and performance on benchmarks like ALFWorld and WebShop.

GenEx is a novel graph-based pipeline that converts SARS-CoV-2 gene sequences into codon co-occurrence graphs to detect variants, using techniques like MSCG and LAPCG, and demonstrates effectiveness with benchmarked ML models.

BERTilda is an explainable framework that tracks topic lifecycles in longitudinal text streams by constructing temporal graphs with similarity and flow signals to detect splits, merges, and other transitions, achieving high agreement rates on annotated datasets.

GraphPO is a novel graph-based reinforcement learning framework that represents rollouts as a directed acyclic graph, merging semantically equivalent reasoning paths to reduce redundant exploration and improve credit assignment for large reasoning models.