INDEX

Explanations

words related to deception or exclusion

oai_token-act-pair · gpt-3.5-turbo

words related to exclusion or being left out

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

 Square

-0.62

 stamp

-0.59

 Daily

-0.57

 fever

-0.57

 Everyday

-0.57

 stamps

-0.57

 itch

-0.56

PAT

-0.56

 House

-0.55

til

-0.55

POSITIVE LOGITS

uded

4.10

uding

3.45

udes

2.66

ude

2.46

uders

2.01

usion

1.85

usions

1.75

usive

1.70

uder

1.62

usional

1.28

Activations Density 0.014%

words related to deception or exclusion

words related to exclusion or being left out

No Comments

No Known Activations