INDEX

Explanations

phrases emphasizing a directive or advice - typically starting with "be."

oai_token-act-pair · gpt-3.5-turbo

the imperative form of "be" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

plings

-0.68

 Strait

-0.62

 reconstruction

-0.61

yip

-0.61

 Taxes

-0.60

 Various

-0.60

Bone

-0.58

 Landing

-0.58

 Bars

-0.58

 vanish

-0.57

POSITIVE LOGITS

 able

1.19

 aware

1.15

 thankful

1.15

 afraid

1.05

 careful

1.04

 mindful

1.03

 grateful

1.01

 confident

1.01

 honest

1.00

 wary

0.98

Activations Density 0.165%

phrases emphasizing a directive or advice - typically starting with "be."

the imperative form of "be" in various contexts

No Comments

No Known Activations