INDEX

Explanations

phrases related to promotions or advancements in job roles

oai_token-act-pair · gpt-3.5-turbo

instances of the word "promoted" or its related forms, indicating elevation or advancement in status

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

opic

-0.80

lv

-0.76

chest

-0.68

 Cage

-0.67

bi

-0.67

adra

-0.66

 Banks

-0.66

bian

-0.65

 Strait

-0.65

 Hayward

-0.65

POSITIVE LOGITS

 promoted

1.17

 promotion

1.06

wcs

0.92

 promotions

0.88

 promotes

0.84

 promoting

0.84

olicy

0.80

 Promotion

0.78

 dissemin

0.76

nesday

0.75

Activations Density 0.010%

phrases related to promotions or advancements in job roles

instances of the word "promoted" or its related forms, indicating elevation or advancement in status

No Comments

No Known Activations