hayulalab/trustmem-hayula-paper

Hayula Research Paper — Hayula Research

Hayula AI Lab

Abstract

We present TRUSTMEM-Hayula, an integrated approach combining TRUSTMEM's trustworthy memory consolidation with EvoTest's evolutionary self-improvement, deployed within the Siyaq context engineering framework. TRUSTMEM (arXiv:2606.25161) introduces a Memory Transition Verifier and preference-guided reinforcement learning that achieves 79% reduction in memory corruption by verifying transitions before consolidation. EvoTest (arXiv:2510.13220, ICLR 2026) introduces an Evolver Agent that analyzes exe

Files

File	Description
paper.md	Full paper (Markdown)
README.md	This model card

Citation

@techreport{hayulalab2026trustmemhayula,
    title={Hayula Research Paper — Hayula Research},
    author={Hayula AI Lab},
    year={2026},
    url={https://huggingface.co/hayulalab/trustmem-hayula-paper}
}

---

hayulalab — Open Source AI Research