INDEX

Explanations

names of people or entities preceded by a title or username

oai_token-act-pair · gpt-3.5-turbo

instances of proper nouns, often related to names or identities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 Reynolds

-0.79

avia

-0.78

SK

-0.75

-0.74

 diesel

-0.73

EV

-0.73

DK

-0.70

OV

-0.70

 Interstellar

-0.70

 viruses

-0.69

POSITIVE LOGITS

Bar

2.64

Bar

2.42

bar

2.30

bar

2.21

 Bars

2.05

BAR

1.98

 bars

1.97

bars

1.75

 Barber

1.64

 Barb

1.44

Activations Density 0.214%

names of people or entities preceded by a title or username

instances of proper nouns, often related to names or identities

No Comments

No Known Activations