INDEX

Explanations

words related to levels, such as threat level or sea level

oai_token-act-pair · gpt-3.5-turbo

references to varying degrees or levels of intensity or impact

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

perse

-0.80

letters

-0.73

oche

-0.69

anson

-0.68

iat

-0.66

pict

-0.66

Europe

-0.65

amas

-0.64

eston

-0.64

icles

-0.64

POSITIVE LOGITS

 level

3.66

 levels

2.60

level

2.59

 Level

2.42

Level

2.25

levels

2.24

 LEVEL

2.21

 Levels

2.00

lvl

1.68

 tier

1.48

Activations Density 0.026%

words related to levels, such as threat level or sea level

references to varying degrees or levels of intensity or impact

No Comments

No Known Activations