INDEX

Explanations

high-ranking political positions and individuals

oai_token-act-pair · gpt-3.5-turbo

references to political figures and titles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

Sensor

-0.68

beh

-0.67

 fingert

-0.67

ickle

-0.66

ensitivity

-0.66

 radius

-0.65

urat

-0.63

itivity

-0.62

 toolbar

-0.61

 prelim

-0.61

POSITIVE LOGITS

 Yugoslavia

0.85

 Desmond

0.77

turned

0.75

 Lyndon

0.72

ãĤ¶

0.71

 Edward

0.69

 Fidel

0.68

 Lowell

0.67

oslov

0.67

 Olympia

0.67

Activations Density 0.179%

high-ranking political positions and individuals

references to political figures and titles

No Comments

No Known Activations