INDEX

Explanations

words related to names of places or people

oai_token-act-pair · gpt-3.5-turbo

names or variants related to individuals, particularly with the letter pattern "il"

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

 elig

-0.80

lished

-0.62

 ultras

-0.60

TRY

-0.60

 bluff

-0.60

 entrants

-0.58

 horr

-0.57

 nightmares

-0.57

$$$$

-0.56

 giveaway

-0.56

POSITIVE LOGITS

igans

1.00

felt

0.79

idan

0.75

kil

0.73

nian

0.73

sworth

0.73

velt

0.73

aku

0.72

zen

0.72

frames

0.71

Activations Density 0.092%

words related to names of places or people

names or variants related to individuals, particularly with the letter pattern "il"

No Comments

No Known Activations