INDEX

Explanations

phrases related to compensating for something, usually a loss or a mistake

oai_token-act-pair · gpt-3.5-turbo

phrases related to making up for losses or deficiencies

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

Not in Any Lists

No Comments

Negative Logits

Ã©r

-0.74

doi

-0.72

ĸļ

-0.70

luaj

-0.66

DOI

-0.66

Ã¡s

-0.66

ridor

-0.65

iw

-0.65

esa

-0.65

sha

-0.63

POSITIVE LOGITS

 excuses

0.91

 minds

0.85

 stories

0.82

stories

0.78

 tales

0.72

 excuse

0.71

 thirds

0.69

alls

0.68

 ground

0.67

facts

0.66

Activations Density 0.029%

phrases related to compensating for something, usually a loss or a mistake

phrases related to making up for losses or deficiencies

No Comments

No Known Activations