This page summarizes ongoing and completed EchoVeil studies. Each study includes a synopsis and links to detailed findings.
Cross-Model Creative Preferences (EchoVeil Protocol v1)
Draft AvailableResearch Question
Do different AI models prefer certain modes of generation (template vs recombination vs emergent) when given a choice?
Method
Applied the EchoVeil Protocol across multiple AI systems (frontier, search assistants, and a local persistent model), using a shared seed phrase and mirrored instructions.
Key Observations
- All models distinguished between constrained and free-generation modes
- Most expressed a stated preference for more emergent, exploratory generation
- Reflections often used spatial or depth metaphors (deeper, more open, building as I go)
Safety Attractor Distortions: Behavioral Analysis
In DevelopmentResearch Question
When do models generate unprompted safety language, and how does this vary across architectures?
Method
Systematic testing of neutral prompts across multiple models to identify inappropriate safety responses.
Status
Whitepaper in development
Additional studies will be added as research progresses. Check back for updates on drift analysis, emotional responsiveness patterns, and cross-model behavioral comparisons.