INDEX

Explanations

mentions of a specific username 'ko'

oai_token-act-pair · gpt-3.5-turbo

repeated instances of a specific character sequence or pattern

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

Doc

-0.67

NCT

-0.64

 Tenth

-0.62

 alternative

-0.62

 Extended

-0.61

 Marriott

-0.60

 Ampl

-0.59

ful

-0.58

 competitor

-0.58

dress

-0.57

POSITIVE LOGITS

ko

4.38

kos

2.00

ka

1.86

KO

1.76

ki

1.64

kin

1.63

kan

1.59

ken

1.57

kov

1.54

ku

1.47

Activations Density 0.005%

mentions of a specific username 'ko'

repeated instances of a specific character sequence or pattern

No Comments

No Known Activations