发表机构
Faculty of Data and Decision Sciences, Technion（以色列理工学院数据与决策科学学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
提出HybridAL，通过监控在线稳定化信号在主动学习中自适应地从重新训练切换到微调，在保持宏F1不劣的同时节省高达49%的训练时间并改善校准性能。
AI 中文摘要
训练策略，即是从零开始重新训练还是从之前的检查点进行微调，是主动学习中一个被忽视的决策变量。我们表明这一选择具有可利用的结构：重新训练在早期轮次最为有用，此时每个批次都能显著重塑标注分布，而一旦模型轨迹稳定，微调则变得更加安全。我们提出HybridAL，一种自适应训练调度方法，它监控在线稳定化信号，并在持续稳定后从重新训练切换到微调。两种互补信号，即谱指数变化$\Delta\alpha$（基于权重）和准确率变化$\Delta$Acc（基于验证集），在时间-校准权衡上覆盖了不同的点。在三个编码器骨干网络和六个文本分类任务（每个任务五个随机种子）上，HybridAL保持端点宏F1在0.010的边际内不劣于重新训练和微调，节省了高达49%的重新训练时间，并恢复了重新训练在负对数似然（NLL）所衡量的校准优势中的相当大一部分。与在预先设定的轮次切换的调度相比，HybridAL在适度的额外成本下获得了更低的NLL，表明依赖轨迹的切换比固定的早期切换提供了更强的时间-校准权衡。
英文摘要
Training strategy, namely whether to retrain from scratch or fine-tune from the previous checkpoint, is an overlooked decision variable in active learning. We show that this choice has exploitable structure: retraining is most useful in early rounds, when each batch can substantially reshape the labeled distribution, while fine-tuning becomes safer once the model trajectory stabilizes. We propose HybridAL, an adaptive training schedule that monitors an online stabilization signal and switches from retraining to fine-tuning after sustained stabilization. Two complementary signals, spectral exponent change $Δα$ (weight-based) and accuracy change $Δ$Acc (validation-based), span different points on the time-calibration trade-off. Across three encoder backbones and six text-classification tasks (five seeds each), HybridAL keeps endpoint macro-F1 non-inferior to retraining and fine-tuning at a 0.010 margin, saves up to 49% of retraining time, and recovers a substantial fraction of retraining's calibration advantage as measured by negative log-likelihood (NLL). Compared with schedules that switch at a pre-committed round, HybridAL obtains lower NLL at moderate additional cost, showing that trajectory-dependent switching provides a stronger time-calibration trade-off than fixed early switching.
CommentsAccepted to EMNLP 2026 Main Conference
