INDEX

Explanations

phrases indicating a statement or opinion on a topic

oai_token-act-pair · gpt-3.5-turbo

phrases that express a statement or assertion

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

xtap

-0.85

¥ŀ

-0.76

hooting

-0.75

asu

-0.74

pes

-0.72

phrine

-0.72

mental

-0.71

 conflic

-0.71

tele

-0.70

Tai

-0.69

POSITIVE LOGITS

 goodbye

1.38

 hello

1.03

 aloud

0.89

 Goodbye

0.87

 farewell

0.78

ysis

0.74

ieu

0.71

 definitively

0.69

 sorry

0.68

ings

0.64

Activations Density 0.049%

phrases indicating a statement or opinion on a topic

phrases that express a statement or assertion

No Comments

No Known Activations