INDEX

Explanations

text related to instructions or procedures

oai_token-act-pair · gpt-3.5-turbo

instances of the word "how"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 MÃ©

-0.63

 receiving

-0.62

 critic

-0.61

ub

-0.61

 civilian

-0.61

 Cortex

-0.61

 crus

-0.60

 grain

-0.59

 ï¿½

-0.58

 Interior

-0.58

POSITIVE LOGITS

soever

1.16

ever

1.01

HCR

0.95

bill

0.93

iculty

0.92

how

0.89

nesota

0.80

much

0.80

ãĤ¼ãĤ¦ãĤ¹

0.79

eries

0.78

Activations Density 0.006%

text related to instructions or procedures

instances of the word "how"

No Comments

No Known Activations