INDEX

Explanations

phrases indicating instructions or requests in a polite manner

oai_token-act-pair · gpt-3.5-turbo

requests or prompts for action

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

-0.75

pires

-0.73

Tok

-0.68

Aust

-0.67

senal

-0.67

MpServer

-0.67

arc

-0.67

é¾

-0.67

byss

-0.66

POSITIVE LOGITS

 note

1.16

 refrain

1.09

 forgive

1.08

 Ignore

1.07

 consider

1.07

 enable

1.07

 excuse

1.01

 beware

1.00

 refer

0.99

 disregard

0.99

Activations Density 0.031%

phrases indicating instructions or requests in a polite manner

requests or prompts for action

No Comments

No Known Activations