INDEX

Explanations

phrases related to legal or authoritative actions and commands

oai_token-act-pair · gpt-3.5-turbo

references to individuals, actions, or relevant dialogue in a narrative context

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

acial

-0.90

acc

-0.83

Ãł

-0.82

achev

-0.81

DAC

-0.78

ì

-0.77

Lat

-0.76

ador

-0.75

juven

-0.75

 basin

-0.74

POSITIVE LOGITS

 Wright

2.62

wright

1.38

 Duffy

1.36

 Witt

1.21

Wr

1.20

 Payne

1.14

 Wichita

1.13

Zel

1.09

 Wilkinson

1.07

 Hernandez

1.07

Activations Density 0.346%

phrases related to legal or authoritative actions and commands

references to individuals, actions, or relevant dialogue in a narrative context

No Comments

No Known Activations