← All parts of this equation

Equation 25 · Part 9 · Probing, Sparse Autoencoders, Patching, and Steering: The Main Interpretability Methods, Compared

=

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,
=

What this part means

The expressions on both sides represent the same quantity under the stated assumptions.

Its job in the formula

The equals sign connects the complete expression on the left with the complete expression on the right. Both sides must have compatible units.

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

An equals sign says that the expression on its left and the expression on its right have the same value under the stated definitions and assumptions.

Open the illustrated equality: what the equals sign claims guide →

Sources cited in the article section

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