INDEX

Explanations

Twitter handles and names

oai_token-act-pair · gpt-3.5-turbo

specific sequences of letters and symbols commonly associated with social media handles or usernames

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

ÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤ

-0.71

MFT

-0.69

 conduc

-0.65

 Levant

-0.65

 conclud

-0.62

 Princ

-0.62

CLASSIFIED

-0.60

wcsstore

-0.60

WAYS

-0.59

:=

-0.58

POSITIVE LOGITS

bh

0.92

kk

0.82

qq

0.82

zek

0.81

ua

0.81

zn

0.81

oi

0.80

zu

0.79

OY

0.79

xus

0.79

Activations Density 0.072%

Twitter handles and names

specific sequences of letters and symbols commonly associated with social media handles or usernames

No Comments

No Known Activations