INDEX

Explanations

words related to damage or distress

oai_token-act-pair · gpt-3.5-turbo

various forms of the word "bleed" or its related concepts

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 occupations

-0.58

 merch

-0.57

 Uran

-0.57

aur

-0.56

 Election

-0.56

Model

-0.56

 mail

-0.56

 Active

-0.56

 Meta

-0.55

 proposition

-0.55

POSITIVE LOGITS

bled

4.14

bling

3.91

bles

2.99

ble

2.64

bler

2.59

bly

2.51

blers

2.37

BLE

1.94

bley

1.44

bl

1.35

Activations Density 0.012%

words related to damage or distress

various forms of the word "bleed" or its related concepts

No Comments

No Known Activations