One reason the current public work can be hard to read from the outside is that different readers encounter different entry points. Some begin with a paper. Others start with the repository. Others first encounter the work through media framing about user-side safety. The structure becomes clearer if the branch is read in order.
Layer One: Context as Structure
CXC-7 treats conversational context as a structured object rather than as a vague backdrop. This matters because without a structured view of context, there is no stable way to talk about how risk accumulates across turns.
CXOD-7 extends that logic into contextual offense and defense. It is the layer that asks how contextual dynamics can be destabilized, exploited, or defended against. Together, these two papers establish the formal language underneath the rest of the branch.
Layer Two: A Branch Phenomenon
USCH should be read as one branch phenomenon inside the broader AI context program. It does not stand in for the whole research agenda. It proposes user-side contextual hallucination as a way to describe what can emerge on the user side during prolonged interaction, even when the visible failure is not a single bad message.
This is why the program should not be reduced to one keyword. AI context is the umbrella problem. USCH is one branch that becomes legible within that umbrella.
Layer Three: Post-Interaction Assessment
USCI is the methodological step that follows. It proposes a post-interaction assessment logic through four explicit axes: Factual Reliance, Contextual Awareness, Self-Regulation, and Situational Awareness. The point is not to diagnose a person. The point is to make a user-side interaction trajectory legible in a bounded, inspectable way.
That assessment layer is why the branch can move toward implementation without collapsing directly into runtime moderation claims.
Layer Four: Bounded Execution
A-CSM is the execution layer. It takes the distinctions established by the papers and turns them into a deterministic public-safe pipeline. The public release does not expose everything. It intentionally withholds confidential taxonomy, private evaluation materials, and restricted scoring logic. But it does expose enough to let outside readers inspect what the branch looks like when translated into a real system boundary.
This is why the public-safe core matters. It shows that the research is not only conceptual. It can be translated into a reproducible pipeline while still keeping release discipline.
Why the Branch Matters
The current public branch shows a coherent sequence. It does not ask readers to accept a system without a framework, or a framework without a method, or a method without a release boundary. The branch is staged so that each layer constrains the next.
For media readers, this explains why the work is more than a single paper. For technical readers, it explains why the repository belongs downstream of the papers. For institutional readers, it shows what is public, what is bounded, and what remains outside the current release.