INDEX

Explanations

adjectives conveying danger or risk

oai_token-act-pair · gpt-3.5-turbo

references to dangerous situations or conditions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

gdala

-0.82

olitan

-0.82

ļéĨĴ

-0.80

mination

-0.79

glas

-0.78

hew

-0.78

via

-0.78

roma

-0.77

pel

-0.76

rix

-0.76

POSITIVE LOGITS

 adolesc

0.96

 endanger

0.86

 undermin

0.86

 dangerous

0.80

 danger

0.78

 dangers

0.77

 consequence

0.76

 sounding

0.76

 threats

0.75

 overdose

0.74

Activations Density 0.028%

adjectives conveying danger or risk

references to dangerous situations or conditions

No Comments

No Known Activations