INDEX

Explanations

words that are either misspelled or intentionally altered

oai_token-act-pair · gpt-3.5-turbo

occurrences of the letter 't'

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

ĪĴ

-0.72

 hemor

-0.70

 tremend

-0.70

 destro

-0.70

 behavi

-0.66

 disadvant

-0.64

 convol

-0.63

ħĭ

-0.63

 Palestin

-0.61

 compar

-0.60

POSITIVE LOGITS

urtle

0.92

roph

0.90

ruck

0.90

weet

0.89

itled

0.83

ract

0.82

ween

0.81

otally

0.80

aper

0.78

unes

0.78

Activations Density 0.080%

words that are either misspelled or intentionally altered

occurrences of the letter 't'

No Comments

No Known Activations