INDEX

Explanations

numerical values or statistics

oai_token-act-pair · gpt-3.5-turbo

numerical values, particularly in a specific format or context

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

 Redux

-0.70

 legs

-0.66

 Sisters

-0.60

 mold

-0.60

Son

-0.59

 mine

-0.58

 pussy

-0.58

 pudding

-0.57

 likeness

-0.57

 tails

-0.57

POSITIVE LOGITS

3.28

2.41

2.28

2.11

2.00

1.94

1.82

1.80

1.77

1.76

Activations Density 0.022%

numerical values or statistics

numerical values, particularly in a specific format or context

No Comments

No Known Activations