INDEX

Explanations

document structure indicators, specifically the beginning of various sections or paragraphs

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

 potřeb

-0.40

 dispozici

-0.39

 producciones

-0.37

 informací

-0.37

 acogedor

-0.37

 dinosaurios

-0.36

 construida

-0.35

 freundlichen

-0.35

 asesinado

-0.35

 başka

-0.35

POSITIVE LOGITS

 protoimpl

0.77

Vidite

0.74

PPS

0.73

niſſe

0.73

iſchen

0.71

<unused79>

0.70

 queſto

0.70

<unused28>

0.69

<unused14>

0.69

[@BOS@]

0.69

Activations Density 0.001%

document structure indicators, specifically the beginning of various sections or paragraphs

No Comments

No Known Activations