INDEX

Explanations

words and phrases related to positioning or arrangement in a sequence or line

oai_token-act-pair · gpt-3.5-turbo

references to ordinal numbers and specific categorizations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

mun

-0.72

 ãĤµãĥ¼ãĥĨãĤ£ãĥ¯ãĥ³

-0.67

aded

-0.67

escape

-0.65

anan

-0.62

laim

-0.62

chwitz

-0.61

-0.60

upe

-0.59

atan

-0.59

POSITIVE LOGITS

 Malays

0.79

 equivalents

0.69

 Franks

0.67

DPR

0.66

Ø¯

0.66

Russ

0.66

 counterpart

0.64

 Russians

0.62

 counterparts

0.62

Tokens

0.61

Activations Density 0.994%

words and phrases related to positioning or arrangement in a sequence or line

references to ordinal numbers and specific categorizations

No Comments

No Known Activations