INDEX

Explanations

tweeted messages

oai_token-act-pair · gpt-3.5-turbo

instances of the term "tweeted" and related forms, indicating discussions about social media posts

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

 Binding

-0.63

 initiation

-0.63

 disadvantages

-0.61

 cent

-0.57

 Protector

-0.57

èĢ

-0.56

 shaping

-0.55

Chapter

-0.55

 phase

-0.54

 mold

-0.53

POSITIVE LOGITS

 tweeted

3.55

 retweet

2.36

 tweeting

2.35

 tweets

2.23

 tweet

2.16

 Tweet

1.81

 texted

1.79

 emailed

1.57

Tweet

1.56

Twitter

1.54

Activations Density 0.014%

tweeted messages

instances of the term "tweeted" and related forms, indicating discussions about social media posts

No Comments

No Known Activations