INDEX

Explanations

phrases related to insults

oai_token-act-pair · gpt-3.5-turbo

references to insults and derogatory language

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

negie

-0.73

ills

-0.73

arten

-0.70

enfranch

-0.67

-0.65

frames

-0.64

iggle

-0.64

ulhu

-0.63

olin

-0.63

illon

-0.62

POSITIVE LOGITS

 insult

1.02

 insults

0.96

 insulted

0.95

 disrespect

0.93

 insulting

0.90

 humour

0.83

 caric

0.82

ingly

0.81

 humili

0.80

 hygiene

0.76

Activations Density 0.092%

phrases related to insults

references to insults and derogatory language

No Comments

No Known Activations