Eyla: Toward an Identity-Anchored LLM Architecture with Integrated Biological Priors -- Vision, Implementation Attempt, and Lessons from AI-Assisted Development
Paper • 2604.00009 • Published
Eyla v1 — audit & reproducibility artifacts
Artifacts accompanying "Eyla: Toward an Identity-Anchored LLM Architecture with Integrated Biological Priors" (arXiv:2604.00009):
eyla_transplant_base_fp16.pt— the transplant base checkpoint (fp16)lora_adapters/lora_c3_soul_final.pt— the final "Soul Pass" identity LoRAmodel/— architecture source: SSM block (HiPPO-initialized), brain orchestrator, memory injector, heuristic gatestokenizer/— tokenizer used in the experiments
Published so the paper's claims — including its documented failures — can be audited. Not intended for production use.
Author & research
- 🌐 Website: arifadito.com
- 💻 GitHub: github.com/Adiuk24
- 📄 Eyla identity paper: arXiv:2604.00009
- 🔧 Rust ML training verification tooling: gradient-flow-arbiter
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support