WritePolicyBench: Benchmarking Memory Write Policies under Byte Budgets

2026/01/31 by Edgard El Cham · 1 voice
Computer Science · #cs.PF

paper · pdf · doi:10.48550/arxiv.2602.02574

Abstract

We introduce WritePolicyBench, a benchmark for evaluating memory write policies: decision rules that choose what to store, merge, and evict under a strict byte budget while processing a stream with document/API drift. The benchmark provides (i) task generators with controlled non-stationarity, (ii) an explicit action interface for external memory, (iii) a byte-accurate cost model, and (iv) standardized metrics that measure both task success and budget efficiency.

Citations

• MemoryBank: Enhancing Large Language Models with Long-Term Memory
• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
• Neural Turing Machines
• A survey on concept drift adaptation

Discussions

• Edgard El Cham: WritePolicyBench: Benchmarking Memory Write Policies under Byte Budgets https://arxiv.org/abs/2602.02574 https://arxiv.org/pdf/2602.02574 https://arxiv.org/html/2602.02574 [bsky, 0 points, 0 comments]

Related

• RoMe: Row Granularity Access Memory System for Large Language Models
• Memory Layers at Scale
• Memory-Centric Computing: Solving Computing's Memory Problem
• Practical Code RAG at Scale: Task-Aware Retrieval Design Choices under Compute Budgets
• Benchmarking Memory-Centric Computing Systems: Analysis of Real Processing-in-Memory Hardware