INDEX

Explanations

phrases related to reading, reviewing, and managing books or other written content

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

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

 poffible

-0.69

 reaſon

-0.61

 faſt

-0.59

 Monfieur

-0.59

 ſol

-0.59

 itſelf

-0.57

 ainfi

-0.57

 juſ

-0.57

 themſelves

-0.56

 eenige

-0.55

POSITIVE LOGITS

 estekak

0.47

 crook

0.46

 choked

0.42

 skimming

0.42

 choke

0.42

 fend

0.41

平

0.41

 blew

0.41

 fuss

0.40

pu

0.40

Activations Density 0.044%

phrases related to reading, reviewing, and managing books or other written content

No Comments

No Known Activations