INDEX

Explanations

Twitter handles or usernames preceded by the "@" symbol

oai_token-act-pair · gpt-3.5-turbo

mentions of Twitter handles

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

 pneum

-0.77

 Vale

-0.69

 Yuan

-0.69

 exorc

-0.68

 malaria

-0.66

 Aval

-0.66

 Primordial

-0.65

 Reborn

-0.64

 Dracula

-0.64

 concentrating

-0.64

POSITIVE LOGITS

#$

1.32

realDonaldTrump

1.18

yss

0.96

gmail

0.87

lis

0.82

wik

0.81

groups

0.81

"@

0.80

gary

0.77

bay

0.77

Activations Density 0.020%

Twitter handles or usernames preceded by the "@" symbol

mentions of Twitter handles

No Comments

No Known Activations