INDEX

Explanations

phrases related to boolean values (true or false)

oai_token-act-pair · gpt-3.5-turbo

statements about truth and falsehood

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

 approach

-0.73

 intern

-0.72

 licensing

-0.72

 fees

-0.71

 disposal

-0.70

 cost

-0.70

 review

-0.67

 overhaul

-0.67

 supervision

-0.66

 announcement

-0.66

POSITIVE LOGITS

true

3.57

false

3.55

False

2.02

True

1.91

null

1.90

truth

1.55

nil

1.18

ALSE

1.16

reality

1.14

successful

1.11

Activations Density 0.015%

phrases related to boolean values (true or false)

statements about truth and falsehood

No Comments

No Known Activations