INDEX

Explanations

address-related elements, particularly those that include numerical information

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_20/width_131k/average_l0_114

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 esternos

-0.44

 AppBar

-0.39

gpl

-0.39

 diversity

-0.39

UserScript

-0.37

 createSlice

-0.37

ticides

-0.37

 InputDecoration

-0.37

 complacent

-0.35

 compulsion

-0.34

POSITIVE LOGITS

 nakalista

0.66

 насељу

0.53

 ModelExpression

0.50

northern

0.48

'},

0.47

("")]

0.47

 snippetHide

0.46

OGND

0.46



0.45

 western

0.44

Activations Density 0.005%

address-related elements, particularly those that include numerical information

No Comments

No Known Activations