The evidence is limited to one training corpus, modest token budgets, models no larger than 350M parameters, no throughput measurements, and mostly few seeds; the 350M tier has one seed per arm, and natural incidence of the projection gate’s adversarial direction blind spot was not measured. · CiteArk
Not assessedNo independent reproduction scheduledLimitationc-limitations
The evidence is limited to one training corpus, modest token budgets, models no larger than 350M parameters, no throughput measurements, and mostly few seeds; the 350M tier has one seed per arm, and natural incidence of the projection gate’s adversarial direction blind spot was not measured.
Source: fixed-paper:PDF p. 8, Section 7, Limitations
Reported and observed measurements
No structured measurement is attached to this Claim.
Assessments (0)
No immutable Assessment has been published for this Claim yet.
