INDEX

Explanations

mentions of items or objects being described as "fancy"

oai_token-act-pair · gpt-3.5-turbo

references to the word "fancy."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

etts

-0.78

upon

-0.77

sen

-0.75

scl

-0.72

essee

-0.70

arenthood

-0.68

iland

-0.67

————————————————

-0.66

IRO

-0.66

amaru

-0.65

POSITIVE LOGITS

 fancy

1.23

pants

0.95

 fanc

0.77

 notions

0.74

tail

0.73

 fries

0.71

 nifty

0.70

 dress

0.70

rous

0.70

vier

0.68

Activations Density 0.014%

mentions of items or objects being described as "fancy"

references to the word "fancy."

No Comments

No Known Activations