MILES dynamically expands step-wise memory with learnable selection heads that rerank candidates and guide reasoning, improving LLM test-time performance under limited supervision.
This regularization encourages the model to rely on more robust features and improves generalization to unseen conditions
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
MILES: Modular Instruction Memory with Learnable Selection for Self-Improving LLM Reasoning
MILES dynamically expands step-wise memory with learnable selection heads that rerank candidates and guide reasoning, improving LLM test-time performance under limited supervision.