INDEX

Explanations

phrases related to a specific individual, McConnell

oai_token-act-pair · gpt-3.5-turbo

references to the name Mitch McConnell

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Negative Logits

DragonMagazine

-0.79

istic

-0.65

istics

-0.64

sid

-0.61

 brim

-0.60

 linem

-0.58

fare

-0.58

running

-0.57

eous

-0.57

 mileage

-0.56

POSITIVE LOGITS

 McConnell

1.22

aney

0.87

hari

0.81

ucky

0.80

eneg

0.79

 Daniels

0.77

apes

0.73

ucket

0.72

esty

0.71

arkin

0.71

Activations Density 0.046%

phrases related to a specific individual, McConnell

references to the name Mitch McConnell

No Known Activations