INDEX

Explanations

words related to dishonesty and lack of integrity

oai_token-act-pair · gpt-3.5-turbo

terms related to quantity and size

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

CAP

-0.67

 clinically

-0.66

 ANGEL

-0.65

 paralyzed

-0.65

 schizophren

-0.65

 stricken

-0.63

 injured

-0.62

 halluc

-0.58

 atten

-0.58

 discriminated

-0.58

POSITIVE LOGITS

enty

4.45

esty

1.44

ewski

1.12

arge

1.11

elve

1.07

esome

1.03

ogle

0.99

doms

0.98

 Plenty

0.96

inx

0.93

Activations Density 0.040%

words related to dishonesty and lack of integrity

terms related to quantity and size

No Comments

No Known Activations