INDEX

Explanations

phrases related to confrontations or arguments

oai_token-act-pair · gpt-3.5-turbo

expressions of dark humor or sarcasm

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

²¾

-0.75

hement

-0.73

Orig

-0.70

 assessments

-0.68

ĳå£«

-0.68

ĨĴ

-0.68

erve

-0.67

isans

-0.67

Scope

-0.67

Asset

-0.67

POSITIVE LOGITS

 masturb

1.27

 kissed

1.19

 dunk

1.18

 kiss

1.13

pee

1.13

 shave

1.13

 wear

1.09

 fart

1.08

 poop

1.07

 lick

1.06

Activations Density 0.609%

phrases related to confrontations or arguments

expressions of dark humor or sarcasm

No Comments

No Known Activations