黑盒模型测试时自适应新方法，零额外调用却大幅提效。
Adapting in the Dark: Efficient and Stable Test-Time Adaptation for Black-Box Models
- 用轻量白盒引导模型生成梯度路径，实现黑盒可控调整。
- ImageNet-C上比ViT-B/16高7.1%，比CLIP高3.4%。
- 商用接口下成本仅为ZOO的1/250，支持实时推理。
针对仅通过API访问的黑盒模型，测试时自适应（TTA）仍是未充分探索的挑战。现有方法如事后输出优化适应能力有限，而零阶优化（ZOO）虽能实现输入空间调整，但在无监督设置下面临高查询代价和优化难题。我们提出BETA（Black-box Efficient Test-time Adaptation），通过引入轻量级局部白盒引导模型，构建可操作的梯度路径。结合预测一致性调节、一致性正则化及面向提示学习的过滤机制，BETA在无需额外API调用且推理延迟几乎不变的前提下，实现稳定适应。在ImageNet-C上，相较于ViT-B/16提升7.1%准确率，相较于CLIP提升3.4%，超越强白盒与灰盒方法（如TENT、TPT）。在商用API上，性能接近ZOO但成本低250倍，同时保持实时推理速度，确立了其在真实黑盒场景下的高效实用性。
原文摘要 · Abstract (English)
Test-Time Adaptation (TTA) for black-box models accessible only via APIs remains a largely unexplored challenge. Existing approaches such as post-hoc output refinement offer limited adaptive capacity, while Zeroth-Order Optimization (ZOO) enables input-space adaptation but faces high query costs and optimization challenges in the unsupervised TTA setting. We introduce BETA (Black-box Efficient Test-time Adaptation), a framework that addresses these limitations by employing a lightweight, local white-box steering model to create a tractable gradient pathway. Through a prediction harmonization technique combined with consistency regularization and prompt learning-oriented filtering, BETA enables stable adaptation with no additional API calls and negligible latency beyond standard inference. On ImageNet-C, BETA achieves a +7.1% accuracy gain on ViT-B/16 and +3.4% on CLIP, surpassing strong white-box and gray-box methods including TENT and TPT. On a commercial API, BETA achieves comparable performance to ZOO at 250x lower cost while maintaining real-time inference speed, establishing it as a practical and efficient solution for real-world black-box TTA.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。