INDEX

Explanations

proper nouns, specifically the name "Duncan"

oai_token-act-pair · gpt-3.5-turbo

mentions of the name "Duncan"

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

)=(

-0.76

front

-0.68

 à¤

-0.67

 rally

-0.65

sy

-0.62

rous

-0.62

ction

-0.61

 affili

-0.61

eral

-0.60

Form

-0.59

POSITIVE LOGITS

 Duncan

3.99

Dun

1.92

 Ashton

1.37

 Upton

1.27

 Dudley

1.18

 Parker

1.13

 Crawford

1.11

 Payton

1.11

 Bowen

1.09

Dun

1.08

Activations Density 0.016%

proper nouns, specifically the name "Duncan"

mentions of the name "Duncan"

No Comments

No Known Activations