INDEX

Explanations

phrases prompting the reader to access further information or read additional content

oai_token-act-pair · gpt-3.5-turbo

instances of the word "Read" or related calls to action encouraging further reading

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

anners

-0.61

 anim

-0.59

otions

-0.59

idia

-0.59

acity

-0.59

 lying

-0.59

ackle

-0.58

ountain

-0.58

ukong

-0.56

fo

-0.55

POSITIVE LOGITS

 Read

3.89

Read

2.17

 READ

1.92

read

1.86

 read

1.85

 Write

1.65

READ

1.56

 Reading

1.51

reads

1.49

 reads

1.48

Activations Density 0.015%

phrases prompting the reader to access further information or read additional content

instances of the word "Read" or related calls to action encouraging further reading

No Comments

No Known Activations