INDEX

Explanations

actually

np_max-act · gemini-2.0-flash

This neuron responds to hedging or emphasis adverbs—most prominently “actually” (and similar intensifiers like “really”).

oai_token-act-pair · o4-mini Triggered by @xinyanhu8

New Auto-Interp

Configuration

andyrdt/saes-llama-3.1-8b-instruct/resid_post_layer_11/trainer_1

Dataset (Dashboard)

Various

Features

131,072

Data Type

float32

Hook Name

blocks.11.hook_resid_post

Architecture

standard

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

CompleteListener

-0.07

ī

-0.06

 extending

-0.06

simulate

-0.06

 sitcom

-0.06

cez

-0.06

 flee

-0.06

>();↵↵

-0.06

Tre

-0.06

₂

-0.06

POSITIVE LOGITS

 actually

0.10

&#

0.07

 вищ

0.07

 действительно

0.07

 phản

0.07

**)&

0.06

(STD

0.06

 gerçekten

0.06

음

0.06

.Align

0.06

Activations Density 0.024%

actually

This neuron responds to hedging or emphasis adverbs—most prominently “actually” (and similar intensifiers like “really”).

No Comments

No Known Activations