INDEX

Explanations

phrases related to a strong level of knowledge or understanding

oai_token-act-pair · gpt-3.5-turbo

references to "knowledge."

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

 capped

-0.73

ques

-0.68

 Slim

-0.68

pez

-0.67

 Appears

-0.67

umbered

-0.67

ump

-0.66

Split

-0.66

ervative

-0.64

 headlined

-0.62

POSITIVE LOGITS

 knowledge

3.71

knowledge

2.50

 Knowledge

2.26

 understanding

1.83

 familiarity

1.71

 ignorance

1.69

 expertise

1.67

 wisdom

1.65

 awareness

1.63

 knowing

1.57

Activations Density 0.010%

phrases related to a strong level of knowledge or understanding

references to "knowledge."

No Comments

No Known Activations