I’m a high school student with a strong interest in AI safety and large language models. Over the past few months, I independently worked on a research paper exploring layered safety architectures for LLM systems, including evaluation frameworks, refusal consistency, context degradation, and safety-performance tradeoffs.
I recently completed the paper and companion repository, and I’m currently trying to submit it to arXiv under cs.AI. Since I’m an independent/student researcher, I require an endorsement to submit.
That said, I completely understand that endorsements should be given carefully. If anyone is willing to briefly review the work, provide critique, or suggest improvements instead, I’d genuinely appreciate that as well. I’m very open to feedback and learning.
Link to paper:
https://drive.google.com/file/d/15iR36Hy3iT33wrDTW1i9onrX4-OS5aEv/view?usp=sharing
If you’d be willing to endorse the submission after reviewing it, you can use this link:
https://arxiv.org/auth/endorse?x=X6QIXU
Thank you for your time.