INDEX

Explanations

phrases urging the audience to stay updated for more information or news

oai_token-act-pair · gpt-3.5-turbo

phrases related to staying updated or informed

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

ãĥīãĥ©ãĤ´ãĥ³

-0.81

ãĤ¨ãĥ«

-0.80

inese

-0.74

BA

-0.69

 Paint

-0.68

phot

-0.67

venge

-0.67

ston

-0.65

Kod

-0.65

é¾įå¥ĳå£«

-0.63

POSITIVE LOGITS

 tuned

0.96

reon

0.79

 ahead

0.76

 tuning

0.74

ÃįÃį

0.70

Tun

0.66

ettings

0.66

eful

0.66

 developments

0.65

aturdays

0.65

Activations Density 0.012%

phrases urging the audience to stay updated for more information or news

phrases related to staying updated or informed

No Comments

No Known Activations