# [cs.CL] arXiv endorsement request — production compound-AI persona & memory architecture (independent researcher)

> Source: <https://discuss.huggingface.co/t/cs-cl-arxiv-endorsement-request-production-compound-ai-persona-memory-architecture-independent-researcher/176793#post_1>
> Published: 2026-06-14 19:18:31+00:00

Hi all,

I’m an independent founder/researcher (Light Ark Technologies — no university or big-lab affiliation), looking for a **cs.CL endorser** so I can post my first arXiv preprint. Endorsement is genuinely the hardest gate when you’re unaffiliated, so any help is hugely appreciated.

**Paper:** *ArkOS Hard-Coded Persona Architecture: A Model-Agnostic Framework for Persistent AI Personality*

A five-stage compound-AI pipeline for production conversational systems, addressing three gaps:

The memory engine re-ranks recalled items by recency/salience (motivated by *Lost in the Middle*), and the overall design builds on MemGPT’s “LLM-as-OS” framing. It’s grounded in a real, self-hosted production deployment (vLLM), and is model-agnostic — the writer model was swapped 27B→31B with no changes to the persona definition.

If you’re a registered cs.CL endorser and willing to help — it takes about a minute and is **not** peer review:

→ Go to [Log in to arXiv | arXiv e-print repository](https://arxiv.org/auth/endorse) , enter code **IFQKQR**, and click endorse.

I’m glad to share the full PDF (here or by email) for anyone who’d like to skim it first. Thank you so much!

— Haijun Wen, Light Ark Technologies, Inc.