PromptMemory
Cross-source consensus on PromptMemory from 1 sources and 5 claims.
1 sources · 5 claims
How it works
Benefits
Risks & contraindications
Highlighted claims
- PromptMemory stores context keys, historical best parameters, and learnable soft prompts. — RASP-Tuner: Retrieval-Augmented Soft Prompts for Context-Aware Black-Box Optimization in Non-Stationary Environments
- PromptMemory retrieves the three nearest memory entries by Euclidean distance and converts negative distances into softmax weights. — RASP-Tuner: Retrieval-Augmented Soft Prompts for Context-Aware Black-Box Optimization in Non-Stationary Environments
- Retrieved prompts and historical bests are weighted averages of the selected memory entries. — RASP-Tuner: Retrieval-Augmented Soft Prompts for Context-Aware Black-Box Optimization in Non-Stationary Environments
- Context masking in NoMemory-RASP increased regret, indicating that retrieval encoded cross-regime structure. — RASP-Tuner: Retrieval-Augmented Soft Prompts for Context-Aware Black-Box Optimization in Non-Stationary Environments
- PromptMemory can create privacy risks because it may store context vectors and parameter traces tied to users or infrastructure. — RASP-Tuner: Retrieval-Augmented Soft Prompts for Context-Aware Black-Box Optimization in Non-Stationary Environments