INDEX

Explanations

occurrences of the word "have" in various forms

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

بوابة

-0.60

Portail

-0.53

ductory

-0.53

 betweenstory

-0.52

₂+

-0.46

 purpoſe

-0.45

quelize

-0.45

XmlAccessType

-0.45

̈́

-0.44

تقاوى

-0.42

POSITIVE LOGITS

 been

0.73

 have

0.73

Have

0.73

have

0.67

 Have

0.67

HAVE

0.59

 gotten

0.59

had

0.58

 taken

0.57

 HAVE

0.57

Activations Density 0.020%

occurrences of the word "have" in various forms

No Comments

No Known Activations