INDEX

Explanations

words related to people and political figures

oai_token-act-pair · gpt-3.5-turbo

individual characters or specific sequences of letters

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 contrace

-0.73

 tremend

-0.68

GMT

-0.67

 convol

-0.65

 readable

-0.64

wcs

-0.62

 cumbers

-0.59

 bearer

-0.58

 millenn

-0.57

artifacts

-0.57

POSITIVE LOGITS

azaki

0.82

uala

0.80

enegger

0.77

shi

0.76

iza

0.74

uchi

0.74

ki

0.74

ÑĮ

0.73

ua

0.73

izu

0.73

Activations Density 0.243%

words related to people and political figures

individual characters or specific sequences of letters

No Comments

No Known Activations