INDEX

Explanations

words written using characters from different languages or possibly corrupted text

oai_token-act-pair · gpt-3.5-turbo

symbols or characters related to formatting or encoding issues

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

 Mandela

-0.64

ifying

-0.64

 cloaked

-0.63

icity

-0.62

 LIFE

-0.62

 Perkins

-0.62

 Dragonbound

-0.61

ification

-0.61

"]=>

-0.60

 confounding

-0.59

POSITIVE LOGITS

ĥ

2.17

Ĺ

1.97

ħ

1.88

Ľ

1.81

ī

1.81

Ļ

1.77

ı

1.76

Ŀ

1.75

ĳ

1.75

Į

1.72

Activations Density 0.024%

words written using characters from different languages or possibly corrupted text

symbols or characters related to formatting or encoding issues

No Comments

No Known Activations