Adapting Without Gradients: Affine Statistics Transport and What Its Certificate Can Tell You
- 类型：arxiv
- 标识：2609.00374
- 链接：https://arxiv.org/abs/2609.00374
- 主分类：llm-infra
- 形态：application
- 被引：1
- 被引来源：Semantic Scholar
- S2被引：1
- 影响力被引：0
- TLDR：CASTER is introduced, a gradient-free method that stores source class statistics in a discriminative subspace, estimates a class-shared affine transformation from target-batch moments, and analytically transports the source class distributions before classification, which positions CASTER as a lightweight adaptation mechanism for frozen-model deployment.
- 待LLM分类：否
- 标题中文：无梯度自适应：仿射统计迁移及其证书能告诉你什么
- TLDR中文：本文提出 CASTER，一种无梯度方法：将源类统计量存储在判别子空间中，从目标批次矩估计一个类共享的仿射变换，并在分类前解析地将源类分布迁移到目标域，使其成为面向冻结模型部署的轻量适配机制。
- 来源文件：
- /inbox/tom/_candidates/2026-09-03-agent-rag-longcontext-candidates.json
- [S2 enrich]