INDEX

Explanations

Spooky and mysterious references

oai_token-act-pair · gpt-3.5-turbo

mentions of names or titles associated with individuals, particularly those ending in "ett."

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

ĺħ

-0.87

¥ŀ

-0.84

xual

-0.73

 tremend

-0.72

ħĭ

-0.70

tymology

-0.68

 metic

-0.68

¥µ

-0.67

 lapse

-0.66

Ĥª

-0.64

POSITIVE LOGITS

erer

0.95

iquette

0.93

sburgh

0.88

eland

0.86

ributed

0.85

lements

0.84

ett

0.84

erers

0.81

rain

0.80

rox

0.79

Activations Density 0.021%

Spooky and mysterious references

mentions of names or titles associated with individuals, particularly those ending in "ett."

No Comments

No Known Activations