Research and releases
Research programs, papers, benchmarks, essays, and open artifacts from Daios. Maintained work and dated publications are labeled separately.
| 4 |
Virtue-Based Character Training Against Compartmentalized Harm
Read on Daios
When a harmful goal is split across agents, no worker may see enough to recognize the plan. This study measures that context limit and tests whether character training remains useful when an attacker changes the prompt. Under identical hostile prompts, training improved safety and helpfulness for Mistral at both tested levels and Qwen3-32B at one. Qwen3-8B traded helpfulness for safety, while Gemma changed little. |
research project Daios Research PREPRINT V1.0 | |
| 3 |
Virtue Ethics-Based Character Training: Building Truth-Telling AI
Open external publication
A dated technical log with philosophical commentary from the Parrhesia project. |
technical report Daios Substack working note | |
| 2 |
The Platonic Case Against AI Slop
Open external publication
Why consuming AI-generated content degrades our capacity to recognize genuine quality. Plato's theory of mimesis applied to model collapse. |
essay Palladium final | |
| 1 |
User-Sovereign Values
Open external publication
Post-training models modularly with LoRA adapters tied to user-selected ethical systems, whether cultural, religious, political, or personal. A plurality of worldviews made programmable. Not one alignment for everyone, but alignment as individual choice. |
research project arXiv ONGOING | |
| 0 |
Beyond Bias and Compliance: Towards Individual Agency and Plurality of Ethics in AI
Open external publication
The founding argument. Why AI ethics must center individual agency, not institutional compliance. |
technical report arXiv final |
Sorted by Date, descending.