Open WorldDiscovery GitHub
← Research library
2026arXiv

Remembering Has a Price: Learning Memory Worth under Resource Constraints

What this work does

TokenWorth learns query-dependent memory usefulness from task outcomes while keeping the language-model backbone frozen. The same learned score guides immediate retrieval and admission, replacement, and eviction in a bounded memory bank.

THE RESEARCH CONVERSATION

Discussion & reactions

Share a question, an insight, or a replication experience. Sign in with GitHub to join the conversation.

Open GitHub discussions ↗