INDEX

Explanations

dates and proper names in a structured format

oai_token-act-pair · gpt-3.5-turbo

sequences of numerical data or percentages in discussions

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

 antioxid

-0.72

Magikarp

-0.71

 exha

-0.68

 adolesc

-0.68

-0.66

.""

-0.63

Balt

-0.62

DAQ

-0.62

 undermin

-0.59

 corrid

-0.59

POSITIVE LOGITS

↵

1.68

<|endoftext|>

1.21

↵Âł

1.11

↵↵

1.09

 While

0.79

 These

0.77

 This

0.77

 However

0.75

 Edit

0.75

 Also

0.74

Activations Density 1.815%

dates and proper names in a structured format

sequences of numerical data or percentages in discussions

No Comments

No Known Activations