Tagged articles
Memory Compression
15 articles · Page 1 of 1
Java agents often work in demos but crash in production due to stack mismatches, governance gaps, and runtime issues such as memory drift, large tool outputs, and lack of observability; the Spring‑Ai‑Trip harness adds progressive compression, spill protection, skill injection, hot‑plug tools, and concurrent execution to bridge the gap.
AI AgentJavaMemory Compression
0 likes · 31 min read

MIT researchers introduce Attention Matching, a latent‑space KV‑cache compaction technique that reduces large‑language‑model memory usage up to 50‑fold with negligible precision loss, outperforming token‑pruning, summarization, and prior compaction methods across benchmarks like QuALITY, LongHealth, and AIME‑2025.
Attention MatchingKV CacheLLM
0 likes · 13 min read

The article presents a technical deep‑dive into TencentDB Agent Memory’s short‑term memory compression, which combines context offloading and a Mermaid‑based infinite canvas to reduce token usage by up to 61 % while improving task success rates by over 50 % across multiple long‑session benchmarks.
AgentLLMLong Context
0 likes · 45 min read


AI WakaMay 12, 2026 · Artificial Intelligence
Through ten experiments on three LLMs, this study measures TurboQuant’s 3‑bit KV‑cache compression, revealing that while quality remains strong, speed gains vary by model, memory savings depend on implementation, and attention‑entropy analysis explains why 2‑bit compression degrades performance.
Attention EntropyKV CacheLLM quantization
0 likes · 14 min read

The article explains that AI agents’ forgetting and hallucinations stem from token‑level attention scores causing key‑value cache eviction before retrieval, then surveys KV‑cache basics, naive growth, streaming‑LLM windowing, SnapKV’s attention‑guided compression, token‑retention studies, Memory Sparse Attention, compares these methods, and discusses practical system pitfalls and design implications.
AI agentsAttentionKV Cache
0 likes · 20 min read


SuanNiApr 3, 2026 · Artificial Intelligence
The article presents the GEMS (Agent‑Native Multimodal Generation with Memory and Skills) framework, detailing its multi‑agent loop, hierarchical memory compression, on‑demand skill modules, and extensive benchmark results that show a lightweight 6B model surpassing larger proprietary systems on complex image‑generation tasks.
GEMSImage GenerationMemory Compression
0 likes · 14 min read

This article reviews three hot open‑source projects—TurboQuant Plus for compressing LLM memory, a Claude‑Code plugin that leverages Codex for AI‑driven code review, and the Rust‑based WeCom CLI for terminal control of Enterprise WeChat—detailing their features, usage, and target users.
AI code reviewClaudeLLM
0 likes · 8 min read

TurboQuant, presented at ICLR 2026, introduces a theoretically grounded vector quantization technique that reduces large‑language‑model key‑value cache memory by at least six times, achieves up to eight‑fold speedups, and maintains zero accuracy loss by combining PolarQuant’s polar‑coordinate compression with a 1‑bit QJL error‑correction step, as demonstrated on benchmarks such as LongBench and GloVe.
AI inferenceMemory CompressionTurboQuant
0 likes · 10 min read

Engula is a Redis‑compatible, high‑performance in‑memory database that cuts memory usage by up to 50% through compression and metadata optimization, while incurring only about 10% performance overhead, and its architecture, testing methodology, and benchmark results are detailed in this article.
Memory Compressioncompatibilityin-memory database
0 likes · 7 min read

This article explains the ZRAM technology in Linux, covering its principles, configuration steps, kernel integration, performance optimizations, and practical use cases for improving memory utilization on embedded devices, Android, and legacy PCs.
Memory CompressionSwapkernel
0 likes · 24 min read

This article explains how Linux memory compression techniques such as zSwap, zRAM, and the Zstandard algorithm reduce I/O pressure, extend flash lifespan, and improve overall system performance, while also covering their drawbacks and step‑by‑step activation procedures.
LinuxMemory CompressionZstd
0 likes · 6 min read

Memory compression techniques like Linux's zSwap, zRAM, and the zstd algorithm reduce I/O latency and extend RAM capacity by compressing swap pages, offering performance gains while introducing trade‑offs such as CPU overhead and configuration complexity, and this guide explains their principles, advantages, drawbacks, and activation steps.
LinuxMemory CompressionSystem Performance
0 likes · 6 min read

The article investigates why the Feishu app sometimes fails to respond to swipe gestures after a hot start, tracing the issue to the app entering a D (uninterruptible) state during memory compression, and demonstrates how adjusting CPU priority for compression threads can reduce the problem's occurrence.
AndroidInput EventsMemory Compression
0 likes · 8 min read

This article explains Redis's ziplist compressed list structure, its internal fields, lookup algorithm, performance characteristics, configuration thresholds for Hash and List types, and demonstrates a real‑world use case with memory‑saving calculations and experimental results.
Memory Compressiondata structuresredis
0 likes · 11 min read

Linux reduces RAM pressure through three main compression mechanisms—zSwap, which caches compressed pages before writing to swap; zRAM, a RAM‑backed compressed block device; and zCache, a file‑page compressor—each paired with specialized allocators (zsmalloc, zbud, z3fold) and configurable algorithms, offering trade‑offs in speed, ratio, CPU load, and fragmentation.
LinuxMemory Compressionperformance
0 likes · 12 min read
