INDEX

Explanations

capital letter pairs followed by numbers

oai_token-act-pair · gpt-3.5-turbo

mentions of electronic components or technologies

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

ãĥīãĥ©ãĤ´ãĥ³

-0.97

ãĥĥãĤ¯

-0.80

ãĥ´ãĤ¡

-0.71

tenance

-0.70

itude

-0.69

ãĥīãĥ©

-0.68

noon

-0.66

ãĤ¤ãĥĪ

-0.65

ãĥ³ãĤ¸

-0.65

ãĤ´ãĥ³

-0.64

POSITIVE LOGITS

REG

0.89

entric

0.83

EG

0.83

rowth

0.82

LES

0.81

LER

0.81

ENT

0.80

BIT

0.80

ossip

0.78

regor

0.77

Activations Density 0.007%

capital letter pairs followed by numbers

mentions of electronic components or technologies

No Comments

No Known Activations