Long-context quality probes and KV-cache research on local GPUs: retrieval is not utilization.
100-question 6-dimension long-conversation memory benchmark for Chinese-healthcare AI. Sivon reference: 92/100 mean (2026-05-27).
SOMA: Sovereign Operative Memory Architecture — A cognitive operating system for long-horizon autonomous agents
Long-context stress tests: lost-in-the-middle heatmaps, multi-hop joins, instruction retention, distractors, and the latency and KV-memory bill.
Vectorless agent memory. HTML storage + grep retrieval. LongMemEval-S R@5 = 98.9%. No embedding, no vector DB.
Virtual 1M context gateway for Claude Code with rolling compression, visible model fallback, and fail-closed context limits.
Does your LLM stack hold the plot? A long-horizon context-integrity benchmark: planted facts, locked-decision reversals, namesake bait, and record-first re-entry, on a 10-axis rubric.
Add a description, image, and links to the long-context topic page so that developers can more easily learn about it.
To associate your repository with the long-context topic, visit your repo's landing page and select "manage topics."