INDEX

Explanations

instances of words related to requests or desires

oai_token-act-pair · gpt-3.5-turbo

references to requests or appeals for something

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

dule

-0.67

 Chop

-0.64

 Kelley

-0.64

 Insurance

-0.62

 sshd

-0.62

 Accounting

-0.60

 Frankfurt

-0.60

 Rocket

-0.59

 Junk

-0.59

 Prospect

-0.58

POSITIVE LOGITS

 pleas

1.34

urable

1.32

ured

0.95

ĸļ

0.95

icular

0.94

geoning

0.94

izu

0.88

gon

0.88

uit

0.86

alty

0.84

Activations Density 0.009%

instances of words related to requests or desires

references to requests or appeals for something

No Comments

No Known Activations