INDEX

Explanations

words related to information, investigation, or evidence

oai_token-act-pair · gpt-3.5-turbo

verbs related to communication, responses, and offerings

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

LU

-0.67

ugi

-0.64

sometimes

-0.62

rather

-0.62

loo

-0.62

probably

-0.62

unknown

-0.62

loving

-0.61

not

-0.60

 beware

-0.60

POSITIVE LOGITS

any

1.95

 anything

1.81

ANY

1.47

 anywhere

1.41

nor

1.41

 anyone

1.38

 anybody

1.38

 anymore

1.38

anything

1.26

 slightest

1.22

Activations Density 0.436%

words related to information, investigation, or evidence

verbs related to communication, responses, and offerings

No Comments

No Known Activations