INDEX

Explanations

names related to individuals or figures

oai_token-act-pair · gpt-3.5-turbo

proper nouns, particularly names associated with prominent individuals

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

ysis

-0.96

 Commodore

-0.76

ously

-0.75

 cartridges

-0.74

ĪĴ

-0.72

 shaft

-0.69

 faults

-0.69

Pocket

-0.67

erous

-0.67

room

-0.63

POSITIVE LOGITS

herer

0.93

sels

0.93

het

0.91

kus

0.89

sel

0.87

hee

0.87

byss

0.87

glas

0.86

erva

0.85

sell

0.85

Activations Density 0.093%

names related to individuals or figures

proper nouns, particularly names associated with prominent individuals

No Comments

No Known Activations