INDEX

Explanations

the word "choose" followed by a high activation word or phrase

oai_token-act-pair · gpt-3.5-turbo

expressions related to the act of choosing or making decisions

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

grounds

-0.80

 Compass

-0.77

til

-0.72

anchester

-0.70

soDeliveryDate

-0.69

listed

-0.69

tained

-0.67

lights

-0.63

 Canary

-0.63

 Accessed

-0.61

POSITIVE LOGITS

 make

1.03

 give

1.00

 pursue

1.00

 reproduce

0.97

 create

0.96

 minimize

0.96

 keep

0.96

 defend

0.96

 preserve

0.95

 capitalize

0.95

Activations Density 0.137%

the word "choose" followed by a high activation word or phrase

expressions related to the act of choosing or making decisions

No Comments

No Known Activations