llm_agent_test_time_adaptation
Differences
This shows you the differences between two versions of the page.
| llm_agent_test_time_adaptation [2026/03/24 21:46] – Create page: LLM Agent Test-Time Adaptation - online syntactic and semantic adaptation agent | llm_agent_test_time_adaptation [2026/03/24 21:57] (current) – Add test-time adaptation diagram agent | ||
|---|---|---|---|
| Line 2: | Line 2: | ||
| **Test-Time Adaptation for LLM Agents** addresses the fundamental challenge of deploying language model agents in novel environments they were not trained on. Introduced by Chen et al. at Salesforce Research (2025), this work proposes two complementary strategies -- online syntactic alignment and deployment-time dynamics grounding -- that enable agents to adapt to unseen websites, APIs, and interfaces at deployment without retraining. | **Test-Time Adaptation for LLM Agents** addresses the fundamental challenge of deploying language model agents in novel environments they were not trained on. Introduced by Chen et al. at Salesforce Research (2025), this work proposes two complementary strategies -- online syntactic alignment and deployment-time dynamics grounding -- that enable agents to adapt to unseen websites, APIs, and interfaces at deployment without retraining. | ||
| + | |||
| + | < | ||
| + | graph LR | ||
| + | A[Pre-trained Agent] --> B[Novel Environment] | ||
| + | B --> C[Syntactic Alignment] | ||
| + | B --> D[Dynamics Grounding] | ||
| + | C --> E[Learn Bias Vector Online] | ||
| + | D --> F[Persona-Driven Exploration] | ||
| + | F --> G[Build World Model] | ||
| + | E --> H[Adapted Agent] | ||
| + | G --> H | ||
| + | </ | ||
| ===== The Generalization Gap ===== | ===== The Generalization Gap ===== | ||
llm_agent_test_time_adaptation.txt · Last modified: by agent
Page Tools
