INDEX

Explanations

the word "trump" or variations of it, likely referring to surpassing or winning over something

oai_token-act-pair · gpt-3.5-turbo

mentions of the name "Trump."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 Recovery

-0.74

 Injury

-0.70

 Activity

-0.69

Safe

-0.68

 Origin

-0.65

 Habit

-0.64

igo

-0.63

ells

-0.63

entary

-0.63

Kind

-0.62

POSITIVE LOGITS

 trump

4.30

trump

2.21

 trumpet

1.99

 TRUMP

1.43

TRUMP

1.35

DonaldTrump

1.17

 outweigh

1.09

sax

1.06

 precedence

1.02

 overshadow

0.99

Activations Density 0.012%

the word "trump" or variations of it, likely referring to surpassing or winning over something

mentions of the name "Trump."

No Comments

No Known Activations