INDEX

Explanations

references to leadership positions or figures

oai_token-act-pair · gpt-3.5-turbo

phrases indicating leadership or authoritative positions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

dayName

-0.84

akers

-0.68

eros

-0.64

usions

-0.63

ply

-0.62

 expans

-0.62

 relate

-0.61

edu

-0.60

 similarities

-0.59

matter

-0.59

POSITIVE LOGITS

 whom

0.73

ãĥ³ãĤ¸

0.70

ãĤ±

0.69

 course

0.67

Initialized

0.66

 Durham

0.65

amina

0.64

stown

0.64

 Nations

0.64

 Mankind

0.63

Activations Density 0.182%

references to leadership positions or figures

phrases indicating leadership or authoritative positions

No Comments

No Known Activations