INDEX

Explanations

instances where someone is involved in writing or authoring content

oai_token-act-pair · gpt-3.5-turbo

instances of the word "wrote" indicating the action of writing

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 slippery

-0.74

 escal

-0.65

gest

-0.59

match

-0.59

sect

-0.58

 shuttle

-0.57

 cent

-0.57

 dusk

-0.56

 aerial

-0.55

 shades

-0.55

POSITIVE LOGITS

 wrote

3.42

 writes

2.54

wrote

2.49

 write

1.99

 penned

1.86

 writing

1.55

Writing

1.54

Write

1.53

 authored

1.52

write

1.48

Activations Density 0.021%

instances where someone is involved in writing or authoring content

instances of the word "wrote" indicating the action of writing

No Comments

No Known Activations