← All parts of this equation

Equation 25 · Part 2 · Probing, Sparse Autoencoders, Patching, and Steering: The Main Interpretability Methods, Compared

Symbol P

vℓ=1∣P∣∑i∈Phℓ(xi+)  −  1∣N∣∑i∈Nhℓ(xi−),hℓ←hℓ+α vℓ,v_\ell = \frac{1}{|P|}\sum_{i \in P} h_\ell(x_i^{+}) \;-\; \frac{1}{|N|}\sum_{i \in N} h_\ell(x_i^{-}), \qquad h_\ell \leftarrow h_\ell + \alpha\, v_\ell,
PP

What this part means

P occurs below the fraction bar. The quantity above the bar is divided by this expression; zero is excluded as a denominator.

Its job in the formula

P occurs below the fraction bar. The quantity above the bar is divided by this expression; zero is excluded as a denominator.

The passage around this formula

The construction is a mean-difference direction, added back into the residual stream with a tunable strength at inference time: vℓ=1∣P∣∑i∈Phℓ(xi+)  −  1∣N∣∑i∈Nhℓ(xi−),hℓ←hℓ+α vℓv_\ell = \frac{1}{|P|}\sum_{i \in P} h_\ell(x_i^{+}) \;-\; \frac{1}{|N|}\sum_{i \in N} h_\ell(x_i^{-}), \qquad h_\ell \leftarrow h_\ell + \alpha\, v_\ell. where P and N index matched positive and negative examples of the target behaviour, hℓh_\ell is the residual-stream activation at layer ℓ\ell , and α\alpha is a coefficient the operator sets by hand. Nothing in this construction requires understanding why the network represents the concept along this direction, only that adding it produces the intended behavioural shift.

Read this part in the article →

Learn the underlying idea

A variable is a named place for a value. Its letter is a local label: x can mean position in one formula and a data point in another.

Open the illustrated variables: a letter stands for a value guide →

See this notation across published equations →

Sources cited in the article section

These citations provide research context; check each source for the exact claim it supports.