INDEX

Explanations

mentions of Twitter activity in the form of tweets

oai_token-act-pair · gpt-3.5-turbo

instances of the word "tweeted"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

cised

-0.79

phal

-0.75

arist

-0.75

cum

-0.73

pure

-0.71

psy

-0.70

å§«

-0.68

esan

-0.65

thin

-0.64

circ

-0.64

POSITIVE LOGITS

 tweeted

1.02

 tweets

1.01

 tweet

0.98

 tweeting

0.91

 Tweet

0.89

 hasht

0.88

storms

0.87

 hashtag

0.85

Tweet

0.81

storm

0.80

Activations Density 0.009%

mentions of Twitter activity in the form of tweets

instances of the word "tweeted"

No Comments

No Known Activations