INDEX

Explanations

mentions of the term "chan" with varying levels of specificity

oai_token-act-pair · gpt-3.5-turbo

references to the online community '4chan'

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

ences

-0.81

 COUR

-0.76

ENC

-0.75

 Mead

-0.67

 Ferr

-0.65

 Jarrett

-0.63

 Cheong

-0.62

 Moons

-0.62

sson

-0.61

 Luxem

-0.61

POSITIVE LOGITS

chan

1.17

nel

0.94

esthetic

0.90

elist

0.89

bara

0.89

ajor

0.87

icum

0.87

thur

0.87

esan

0.86

hattan

0.85

Activations Density 0.023%

mentions of the term "chan" with varying levels of specificity

references to the online community '4chan'

No Comments

No Known Activations