INDEX

Explanations

terms related to titles or positions of authority, particularly those including the word "Chief"

oai_token-act-pair · gpt-3.5-turbo

references to individuals with the title "Chief"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.4.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ajor

-0.91

AAF

-0.83

pez

-0.81

 destro

-0.80

agra

-0.79

 therap

-0.78

ngth

-0.78

ted

-0.76

\/\/

-0.75

vous

-0.74

POSITIVE LOGITS

 Chief

0.98

Chief

0.86

 Petty

0.82

doms

0.81

 Scientist

0.81

 Executive

0.79

 Engineer

0.78

 Constable

0.77

 Prosecutor

0.76

 Officer

0.75

Activations Density 0.013%

terms related to titles or positions of authority, particularly those including the word "Chief"

references to individuals with the title "Chief"

No Comments

No Known Activations