INDEX

Explanations

words related to employment or hiring, especially the term "employees"

oai_token-act-pair · gpt-3.5-turbo

references to employees or staff members

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 dividing

-0.61

 impe

-0.60

 symbol

-0.60

 fracturing

-0.58

 fragment

-0.57

 manufactured

-0.57

trop

-0.56

 Slash

-0.56

 vicious

-0.55

 Compton

-0.55

POSITIVE LOGITS

ees

4.79

ee

2.76

EE

1.58

eed

1.53

eez

1.39

een

1.32

eem

1.28

eeds

1.23

ey

1.23

ez

1.22

Activations Density 0.004%

words related to employment or hiring, especially the term "employees"

references to employees or staff members

No Comments

No Known Activations