INDEX

Explanations

phrases related to tracking or linking information together

oai_token-act-pair · gpt-3.5-turbo

phrases related to betrayal or deceit

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

 Parenthood

-0.71

aukee

-0.69

ison

-0.60

 Newly

-0.60

 Motorsport

-0.60

Kiw

-0.60

ISON

-0.59

 Blues

-0.58

 constitutionally

-0.58

awan

-0.57

POSITIVE LOGITS

stab

1.33

dated

1.27

tracking

1.26

doors

1.13

packs

1.11

stories

1.10

handed

1.10

fired

1.10

ped

1.09

track

1.08

Activations Density 0.040%

phrases related to tracking or linking information together

phrases related to betrayal or deceit

No Comments

No Known Activations