← All parts of this equation

Equation 27 · Part 5 · A Chatbot Confessed to Being Built by a Company That Never Trained It

Symbol r_m

μc=1∣Mc+∣∑m∈Mc+1[d(rm,ρ)>δ]\mu_c=\frac{1}{|\mathcal{M}_c^+|}\sum_{m\in\mathcal{M}_c^+}\mathbb{1}[d(r_m,\rho)>\delta]
rmr_m

What this part means

m ’s response to probe q.

Its job in the formula

rmr_m is one of the signed contributions combined to compute the quantity on the left.

Where the article explains it

For each model m , define its trait incidence as I(m)=1k\frac{1}{k}∑q∈Q\sum_{q\in Q}1\mathbb{1}[rm(q)r_m(q)∈\inT\mathcal{T}] , where Q is the held-out probe set of size k and rm(q)r_m(q) is m ’s response to probe q .

The passage around this formula

A transfer rate above zero says a channel is associated with a trait. It does not yet say whether the trait is a faithful copy or a garbled one, and it says nothing about whether curators along the way made the trait more or less likely to survive into the next training corpus. Two further quantities complete the construction. The mutation rate μc\mu_c=1∣Mc+∣\frac{1}{|\mathcal{M}_c^+|}∑m∈Mc+\sum_{m\in\mathcal{M}_c^+}1\mathbb{1}[d(rmr_m,ρ\rho)>δ\delta] measures, among the trait-positive models in class c (the set Mc+\mathcal{M}_c^+ ), what fraction produce a response diverging from the seeded canonical form ρ\rho by more than a fixed threshold δ\delta under a declared distance function d : a high μc\mu_c says the…

Read this part in the article →

Learn the underlying idea

A subscript is a label attached below a symbol. It often selects a time step, component, category, or member of a sequence.

Open the illustrated subscripts: which member of a family? guide →

See this notation across published equations →

The article lists its research sources here.