INDEX

Explanations

words related to finality or negativity, such as "dead end" or "dead heat."

oai_token-act-pair · gpt-3.5-turbo

the word "dead" and its variations in various contexts

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

orney

-0.80

ĸļ

-0.74

ECA

-0.71

ãĥ´ãĤ¡

-0.70

Catalog

-0.69

fty

-0.68

ATIONS

-0.68

omal

-0.67

zzle

-0.67

idate

-0.66

POSITIVE LOGITS

beat

1.03

weight

0.93

liest

0.90

lier

0.90

locked

0.89

heading

0.86

spin

0.85

 carc

0.83

lift

0.82

dead

0.80

Activations Density 0.019%

words related to finality or negativity, such as "dead end" or "dead heat."

the word "dead" and its variations in various contexts

No Comments

No Known Activations