argumentA line of reasoning from premises to a conclusion. Like a claim, it belongs to no one who makes it.constitution →
The argument from the inductive bias of interpretable sparse bases
forThis argument, if it holds, bears in favour of the claim.constitution →Sparse probes on sufficiently good sparse autoencoder latents are less likely than dense probes to overfit spurious correlations.
The argument
Its written form has not been written yet.
Does it work?
- Does it follow?
- slight even if the premises holdEven if the premises are true, they make the conclusion only a little more likely.constitution →
- Do the premises hold?
- 1 not separately assessed
Granting that the target concept sits in a few latents, the conclusion still needs the spurious cue to be less cleanly isolated than the concept. Where spurious cues are as well represented by latents, a sparse probe has no reason to prefer the true signal, so the inductive bias favours the claim only a little.
Premises
- UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →Spurious correlates of a label are as likely as the target concept to be represented by individual sparse autoencoder latents.
Where it is made
Passages where sources make, criticize or mention this argument. They record where it appears, not how strong it is, and none has precedence over another.
in hindsight, we're now a lot more confused about whether even an optimal SAE should beat linear probes on OOD generalisation.