INDEX

Explanations

words related to censored or sensitive content

oai_token-act-pair · gpt-3.5-turbo

instances of the sequence "ens" in various forms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 McAuliffe

-0.62

 Bezos

-0.61

 snail

-0.60

vice

-0.59

 Scotia

-0.59

swick

-0.56

 Takeru

-0.56

 Epstein

-0.56

 Doodle

-0.55

 Mata

-0.55

POSITIVE LOGITS

urable

0.97

ource

0.96

hift

0.94

orship

0.94

chen

0.93

haw

0.90

manship

0.85

umer

0.85

ensical

0.84

ured

0.83

Activations Density 0.042%

words related to censored or sensitive content

instances of the sequence "ens" in various forms

No Comments

No Known Activations