INDEX

Explanations

words related to recognition or special acknowledgment

oai_token-act-pair · gpt-3.5-turbo

references to awards and honors

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

 Tube

-0.84

Â£

-0.67

Alternatively

-0.64

imet

-0.64

yip

-0.64

 bowel

-0.63

BN

-0.62

axis

-0.62

Tube

-0.62

 fuelled

-0.61

POSITIVE LOGITS

 honors

4.14

 honor

2.12

 honoring

1.69

 accol

1.56

 honored

1.50

 honorable

1.46

hon

1.39

 privileges

1.38

 honour

1.36

 Honor

1.34

Activations Density 0.009%

words related to recognition or special acknowledgment

references to awards and honors

No Comments

No Known Activations