INDEX

Explanations

people with titles of authority or leadership

oai_token-act-pair · gpt-3.5-turbo

references to individuals in positions of management or editorial leadership

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

çīĪ

-0.68

PU

-0.68

oho

-0.67

qt

-0.67

icipated

-0.65

ppa

-0.65

reek

-0.65

ingers

-0.64

olid

-0.62

acher

-0.62

POSITIVE LOGITS

 managing

0.92

 Managing

0.83

 administrator

0.82

 manage

0.81

 manages

0.79

ahime

0.76

urable

0.75

 hemor

0.70

 manag

0.68

ategy

0.66

Activations Density 0.007%

people with titles of authority or leadership

references to individuals in positions of management or editorial leadership

No Comments

No Known Activations