INDEX

Explanations

phrases related to expressing opinions or statements

oai_token-act-pair · gpt-3.5-turbo

phrases indicating speech or conversation, particularly the verbs "say" and "said"

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

oling

-0.78

oled

-0.78

ï¸ı

-0.73

=~=~

-0.73

blem

-0.72

quartered

-0.70

aliation

-0.69

 Lauder

-0.68

airs

-0.67

isters

-0.67

POSITIVE LOGITS

 goodbye

1.10

bye

0.96

 earlier

0.74

 aloud

0.74

 publicly

0.69

 unto

0.68

 farewell

0.67

 plainly

0.65

 amen

0.63

 bluntly

0.62

Activations Density 0.097%

phrases related to expressing opinions or statements

phrases indicating speech or conversation, particularly the verbs "say" and "said"

No Comments

No Known Activations