Decoupling Memory from Context: Structured Memory for Token-Efficient Test-Time Continual Learning
cs.AI, cs.LG
Submitted: 2026-10-02
Updated: 2026-10-06
Comments: 14, 4, neurips workshop: TTCL
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning
- Doc-to-LoRA: Learning to Instantly Internalize Contexts
- Sweeping Heterogeneity with Smart MoPs: Mixture of Prompts for LLM Task Adaptation
- RIZZ: Routing Interactions to Near Zero-Interference Zones for Continual Adaptation of Black-Box Agents
- LoRA: Low-Rank Adaptation of Large Language Models
- HAGE: Harnessing Agentic Memory via RL-Driven Weighted Graph Evolution
- DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines
- Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills
- Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs
- Dynamic Cheatsheet: Test-Time Learning with Adaptive Memory
- WikiSkill: Compiling Agent Experience into Persistent Knowledge for Skill Evolution
- FinLoRA: Benchmarking LoRA Methods for Fine-Tuning LLMs on Financial Datasets
- Large Language Models as Optimizers
- TextGrad: Automatic "Differentiation" via Text
- Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Agents' Last Exam
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems