INDEX

Explanations

mentions of family relationships, particularly uncles

oai_token-act-pair · gpt-3.5-turbo

repeated references to the word "uncle."

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

 largeDownload

-0.81

*/(

-0.80

mberg

-0.74

RH

-0.70

anchester

-0.69

PsyNetMessage

-0.67

pire

-0.67

ĸļ

-0.65

Draft

-0.65

 Featured

-0.65

POSITIVE LOGITS

 uncle

1.43

 aunt

1.08

 nephew

0.95

 Uncle

0.94

 niece

0.89

liest

0.86

liness

0.83

 Aunt

0.77

Dad

0.75

heses

0.74

Activations Density 0.006%

mentions of family relationships, particularly uncles

repeated references to the word "uncle."

No Comments

No Known Activations