INDEX

Explanations

email-related terms or prompts

oai_token-act-pair · gpt-3.5-turbo

references to email alerts and news notifications

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

fig

-0.89

stood

-0.85

jury

-0.84

flies

-0.81

shr

-0.76

pex

-0.73

ball

-0.71

itar

-0.70

cloth

-0.69

fulness

-0.68

POSITIVE LOGITS

 Emails

1.30

 Transcript

0.91

VERTISEMENT

0.88

 Email

0.79

querque

0.77

 Timeline

0.76

ļéĨĴ

0.76

Typ

0.76

CLASSIFIED

0.72

usalem

0.72

Activations Density 0.006%

email-related terms or prompts

references to email alerts and news notifications

No Comments

No Known Activations