INDEX

Explanations

inappropriate content such as vulgar language

oai_token-act-pair · gpt-3.5-turbo

instances of vulgar language

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

Winged

-0.73

atform

-0.70

ulin

-0.64

eger

-0.63

aah

-0.63

umbledore

-0.62

DOC

-0.61

UL

-0.61

WIND

-0.60

ulation

-0.60

POSITIVE LOGITS

folk

0.77

eric

0.72

lists

0.71

 Strait

0.69

yang

0.65

pend

0.64

 trader

0.63

roth

0.62

 cousin

0.62

shit

0.62

Activations Density 0.000%

inappropriate content such as vulgar language

instances of vulgar language

No Comments

No Known Activations