INDEX

Explanations

the word "subject"

oai_token-act-pair · gpt-3.5-turbo

phrases related to subjects of documentaries or media coverage

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

frog

-0.67

fml

-0.66

pt

-0.64

 condu

-0.62

MX

-0.62

ibr

-0.62

ept

-0.62

gob

-0.60

cup

-0.60

ptions

-0.60

POSITIVE LOGITS

olls

0.71

international

0.71

 ridicule

0.65

 neutrality

0.61

æĺ¯

0.60

 geopolitical

0.60

 international

0.59

 litigation

0.59

ãĥĩ

0.58

 sear

0.58

Activations Density 0.096%

the word "subject"

phrases related to subjects of documentaries or media coverage

No Comments

No Known Activations