INDEX

Explanations

information related to news articles or headlines

oai_token-act-pair · gpt-3.5-turbo

colloquial expressions and punctuation that denote questions or statements

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 util

-0.94

foc

-0.83

 progressing

-0.79

 enqu

-0.79

 redes

-0.79

 endeavour

-0.78

 synerg

-0.78

 advoc

-0.77

 endeav

-0.76

 flavours

-0.75

POSITIVE LOGITS

Romney

1.39

Democrats

1.36

TRUMP

1.36

Hillary

1.35

Republicans

1.34

Obama

1.32

Donald

1.28

CNN

1.27

Trump

1.26

Enlarge

1.23

Activations Density 0.624%

information related to news articles or headlines

colloquial expressions and punctuation that denote questions or statements

No Comments

No Known Activations