InfiniteMem – deterministic episodic memory, LoCoMo 70.49

InfiniteMem, a deterministic, self-hosted episodic memory system for AI agents, reports a LoCoMo QA score of 70.49% and claims a capacity of 1 million episodes or 145 million tokens per deployment, roughly 1,130 times a GPT-4o context window. The system, which is closed source and uses an MCP server for integration, is detailed in a repository by marcot3ssar1, with evaluation scripts and results available for reproducibility.

Deterministic, self-hosted episodic memory for agents K3 engine : 1M episodes / 145M tokens per deployment — ~1,450 novels, ~1,130× a GPT-4o context window, ~27 years of chat. Details: CAPACITY.md /marcot3ssar1/InfiniteMemOs/blob/main/CAPACITY.md . LoCoMo QA 70.49% under equal protocol, reproducible evaluation glue. No implementation sources are included by design. Closed source — see NOTICE. Try it: an MCP server with a free trial is coming — run infiniteMem from Claude Desktop and your agents without sharing data externally. Contact via repository for trial access. paper/ — scientific report v2 LoCoMo + LongMemEval, stronger-reader rows, world-model gate, reranker study, limits, references . harness/ — evaluation scripts only dataset fetch checks, fresh embedding via NVIDIA API, numpy retrieval crosscheck, QA answer+judge via OpenRouter, rerank gate . They drive the closed engine as a black box and reveal nothing about its internals. results/ — certified numbers: leaderboard.json LoCoMo, SOTA2-ready , leaderboard lmem.json LongMemEval-S , readers.json stronger-reader rows + world-model gate , per-question qa cache.jsonl 3070 records, no gold texts , retrieval and QA reports, reranker negative-result report. - Fetch snap-research/locomo data/locomo10.json CC BY-NC 4.0, non-commercial, cite Maharana et al., ACL 2024 and verify SHA256 79FA87E9… . - Set env keys never commit : NVIDIA API KEY embeddings llama-nemotron-embed-vl-1b-v2 , 768-D, passage Speaker: text / query raw and OPENROUTER API KEY answer+judge openai/gpt-4o-mini-2024-07-18 , temp 0, verbatim Mem0 prompts . - Run harness in order: exnovo embed.py → eval fresh retrieval.py → exnovo qa.py --mode k3 + --mode oracle . Expected: retrieval 78.85/74.10 , QA k3 70.49 / oracle 84.89 ±0.5pp API variance . Black-box engine, brute top-10 single-shot, dedup off, no rerank, 1 run. Answer/judge models and prompts pinned see paper §3, §6 . Dataset CC BY-NC 4.0. Contact via repository for engine access.