INDEX

Explanations

sources or citations indicated by the word "via"

oai_token-act-pair · gpt-3.5-turbo

references to sources or citations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

$,

-0.77

 frogs

-0.76

 footing

-0.76

naire

-0.75

$.

-0.75

inary

-0.71

 mornings

-0.67

arse

-0.67

ucket

-0.67

cas

-0.66

POSITIVE LOGITS

 Wikimedia

1.01

Via

0.95

ulture

0.85

WS

0.83

 REUTERS

0.81

 Email

0.79

Via

0.76

via

0.76

 Shutterstock

0.75

VT

0.74

Activations Density 0.012%

sources or citations indicated by the word "via"

references to sources or citations

No Comments

No Known Activations