INDEX

Explanations

names of people, especially in authoritative or legal roles

oai_token-act-pair · gpt-3.5-turbo

official titles and positions related to the judiciary and law enforcement

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 selfies

-0.81

iphate

-0.78

 sucker

-0.74

 homemade

-0.73

 pancakes

-0.71

crazy

-0.69

 ensl

-0.69

venant

-0.68

 revenge

-0.67

ãĤ¨ãĥ«

-0.66

POSITIVE LOGITS

 Deputy

1.69

 Assistant

1.66

 Chief

1.56

 Director

1.50

 Acting

1.49

 Attorney

1.49

 deputy

1.43

 Commissioner

1.42

 Associate

1.41

 Chair

1.37

Activations Density 0.377%

names of people, especially in authoritative or legal roles

official titles and positions related to the judiciary and law enforcement

No Comments

No Known Activations