INDEX

Explanations

unstructured posts containing numbers and personal details

oai_token-act-pair · gpt-3.5-turbo

numerical data and references related to posts or social media metrics

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

ãĤ±

-0.86

ãĥ´

-0.83

ãĤ½

-0.76

 Lauder

-0.72

ãĥĭ

-0.70

numbered

-0.67

ãĥīãĥ©ãĤ´ãĥ³

-0.67

ĸļå£«

-0.67

ãĤ¼

-0.65

ãĥĹ

-0.65

POSITIVE LOGITS

 Quote

0.79

 Seems

0.78

 Edited

0.73

 Removed

0.69

 Interesting

0.66

 Alright

0.66

 Related

0.65

Hmm

0.64

 Thanks

0.61

 moot

0.61

Activations Density 0.037%

unstructured posts containing numbers and personal details

numerical data and references related to posts or social media metrics

No Comments

No Known Activations