01
Fixed data
A fixed real corpus can be read through frozen embedders to produce structured supervision without feeding generated samples back into the source set.
Chris Royse / leapable.ai
Research brief - Submission #64 - OpenReview mpQXCwkQcq
The research brief.
The paper explains Derived Data Abundance, meaning compression, Teleological Constellation Training, and the limits of each claim.
02
Signal ceiling
At N=13, the identity yields up to 91 supervisory signals per input as a constructive upper bound. EXP-2 is still pending.
03
Limit
DDA is a scope argument outside the Shumailov recursion regime. It is not a refutation of model collapse.
The paper is part research artifact, part audit trail.
Watch on YouTube· 3:45
Multi-agent research workflow and audit trail
This video shows the research-production workflow: multiple agents, shared memory, source collection, and validation loops. It is here to explain the operating discipline around the paper; the paper's claims still live or die by the public submission, /measured, /method, and /artifacts.
Watch videoOpen the full video on YouTubeHave one context problem?
Send the audience, data type, target task, proof bar, and sharing limits. The first pass should stay bounded.