INDEX

Explanations

phrases related to causing harm or damage to others

oai_token-act-pair · gpt-3.5-turbo

words or phrases that indicate negative impacts or harm to individuals or communities

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

uther

-0.74

alled

-0.74

aut

-0.73

iliary

-0.70

ult

-0.70

ulum

-0.68

au

-0.68

ials

-0.68

Nap

-0.67

ault

-0.67

POSITIVE LOGITS

 hurting

1.11

 disadvant

0.98

 adolesc

0.92

 harming

0.91

 undermin

0.85

 badly

0.85

 horribly

0.83

lehem

0.82

Pwr

0.81

 harmed

0.80

Activations Density 0.010%

phrases related to causing harm or damage to others

words or phrases that indicate negative impacts or harm to individuals or communities

No Comments

No Known Activations