Postgres memory layer for agents reaches 73.6% on full LongMemEval, not a sampled subset
MycoBrainAI · reddit · 2026-07-21
A Postgres-based memory layer for agents scored 73.6% QA accuracy on the full 500-question LongMemEval oracle set, not a sampled subset.
The builder says the full run exposed problems that the sampled benchmark hid. The biggest improvements came from three choices: handling contradictions explicitly, refusing to store low-confidence extractions, and combining full-text with semantic retrieval instead of using only one. A one-command repro and harness were published in the repo.
More from coding & agent
- A better path to agent autonomy is running waves, finding friction, and iterating — JnBrymn · 2026-07-22
- Coding agents are heading toward an AI-writes, AI-reviews, human-approves workflow — aftahi_ai · 2026-07-22
- oMLX 0.5.2 adds Mac menu-bar stats, low-bit decode kernels, and faster downloads — awnihannun · 2026-07-22
- GitHub review bot hits its PR limit and forces a 39-minute cooldown — DanielLockyer · 2026-07-22
- Max reasoning effort appears to be mobile-only in Codex Remote, not desktop — GabGarrett · 2026-07-22
- A Reddit demo argues online stores should expose carts and pricing through MCP — gelembjuk · 2026-07-22