INDEX

Explanations

phrases where someone is admitting to something

oai_token-act-pair · gpt-3.5-turbo

instances of the word "admit."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

ammy

-0.85

srf

-0.77

ccording

-0.76

 captcha

-0.70

 chimpan

-0.69

 impacted

-0.69

pleting

-0.68

miah

-0.66

 clipboard

-0.66

 tremend

-0.66

POSITIVE LOGITS

ibility

1.05

 admit

0.96

 admits

0.82

 admitting

0.81

essors

0.80

iary

0.79

 admitted

0.78

IBLE

0.78

orial

0.76

glers

0.75

Activations Density 0.014%

phrases where someone is admitting to something

instances of the word "admit."

No Comments

No Known Activations