INDEX

Explanations

words related to hyphenated phrases or compounds

oai_token-act-pair · gpt-3.5-turbo

negative phrases or words related to risks or problematic situations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 Dickinson

-0.83

ulhu

-0.80

AVG

-0.67

 Compass

-0.65

ħĭ

-0.62

 Norris

-0.62

 "$:/

-0.62

 Richards

-0.61

 EntityItem

-0.60

 Loll

-0.60

POSITIVE LOGITS

based

1.17

sized

1.14

level

1.04

shaped

0.95

related

0.95

eyed

0.94

themed

0.92

induced

0.92

centered

0.91

specific

0.91

Activations Density 0.510%

words related to hyphenated phrases or compounds

negative phrases or words related to risks or problematic situations

No Comments

No Known Activations