INDEX

Explanations

the word "Se" followed by a number

oai_token-act-pair · gpt-3.5-turbo

instances of the word "Se."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

CPC

-0.80

 wire

-0.77

Downloadha

-0.75

 anecd

-0.68

 hoop

-0.67

FAA

-0.66

 bulk

-0.66

 deed

-0.65

 overhead

-0.65

SHIP

-0.65

POSITIVE LOGITS

vere

1.20

Se

1.17

eker

1.11

venth

1.04

ldom

1.02

lected

0.97

quel

0.95

lys

0.94

hn

0.93

eme

0.92

Activations Density 0.005%

the word "Se" followed by a number

instances of the word "Se."

No Comments

No Known Activations