INDEX

Explanations

a keyword related to prohibition or restriction

oai_token-act-pair · gpt-3.5-turbo

instances of the word "banned."

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

Â¯Â¯

-0.77

sis

-0.71

ensional

-0.70

eah

-0.68

cknowled

-0.66

slave

-0.65

eworks

-0.65

Barn

-0.65

Appearance

-0.64

Cloud

-0.64

POSITIVE LOGITS

 banned

1.19

 banning

1.02

 bans

0.93

 prohibited

0.90

wana

0.87

 suspending

0.87

 outlaw

0.86

ban

0.85

 cens

0.83

 prohibition

0.81

Activations Density 0.010%

a keyword related to prohibition or restriction

instances of the word "banned."

No Comments

No Known Activations