INDEX

Explanations

phrases involving coordinating conjunctions such as "and"

oai_token-act-pair · gpt-3.5-turbo

conjunctions that connect thoughts or ideas

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 millenn

-0.68

 Azerb

-0.55

Fed

-0.55

 Palestin

-0.54

Magikarp

-0.54

 depos

-0.53

ilty

-0.52

 isolation

-0.52

WORK

-0.52

icity

-0.52

POSITIVE LOGITS

rogen

1.50

rogens

1.42

ro

1.07

then

1.03

romeda

1.02

rew

0.97

rology

0.92

rea

0.89

ros

0.88

rost

0.85

Activations Density 0.173%

phrases involving coordinating conjunctions such as "and"

conjunctions that connect thoughts or ideas

No Comments

No Known Activations