INDEX

Explanations

Pokémon card names

oai_token-act-pair · gpt-3.5-turbo

encoded or non-standard characters in text

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

 Bourbon

-0.76

 Franklin

-0.74

 Toledo

-0.64

 Johns

-0.62

SEC

-0.62

yth

-0.61

 whis

-0.61

Sha

-0.60

 Cord

-0.60

Ole

-0.59

POSITIVE LOGITS

ãĥ¼

4.41

ãĥ³

3.23

ãĥ¼ãĥ

2.80

ãĥĥãĤ¯

2.52

ãĥĥ

2.48

ãĥ«

2.41

ãĥ¼ãĥ«

2.34

ãĥ¼ãĤ¯

2.32

ãĤ¤

2.22

ãĤ¹

2.19

Activations Density 0.020%

Pokémon card names

encoded or non-standard characters in text

No Comments

No Known Activations