INDEX

Explanations

information related to recipes and cooking instructions

oai_token-act-pair · gpt-3.5-turbo

phrases related to recipes or cooking instructions

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 forensic

-0.80

 obstruction

-0.75

 tele

-0.73

 diagn

-0.71

 pursu

-0.70

 debugger

-0.70

 licensing

-0.69

 interference

-0.69

 involuntary

-0.68

 conver

-0.68

POSITIVE LOGITS

Ingredients

1.72

Recipe

1.50

 Ingredients

1.43

 Recipe

1.38

Chicken

1.34

 Serve

1.30

 Delicious

1.29

Nut

1.21

reci

1.17

 Bake

1.16

Activations Density 0.265%

information related to recipes and cooking instructions

phrases related to recipes or cooking instructions

No Comments

No Known Activations