INDEX

Explanations

phrases indicating requests or commands

oai_token-act-pair · gpt-3.5-turbo

expressions of strong emotions or commands

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

arthed

-0.86

 Additional

-0.72

umbn

-0.71

ounter

-0.69

entials

-0.65

è¦ļéĨĴ

-0.65

atana

-0.64

cedented

-0.64

"},"

-0.64

]),

-0.62

POSITIVE LOGITS

 kidding

1.11

 gotta

0.93

ain

0.93

?!

0.89

 fuckin

0.86

hhhh

0.82

 gonna

0.81

ahah

0.81

 yeah

0.80

cknow

0.80

Activations Density 0.364%

phrases indicating requests or commands

expressions of strong emotions or commands

No Comments

No Known Activations