INDEX

Explanations

words related to searching and seeking

oai_token-act-pair · gpt-3.5-turbo

mentions of people or professionals involved in watching or observing

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

 breakup

-0.65

dx

-0.64

 reboot

-0.61

 Rocket

-0.58

 Ross

-0.57

 Psychiatric

-0.56

 grad

-0.56

 relative

-0.56

br

-0.56

OC

-0.55

POSITIVE LOGITS

chers

4.90

cher

3.33

chery

2.56

ches

2.08

ched

1.86

ching

1.71

glers

1.38

che

1.30

chens

1.28

lers

1.24

Activations Density 0.010%

words related to searching and seeking

mentions of people or professionals involved in watching or observing

No Comments

No Known Activations