INDEX

Explanations

the action of physically opening something or initiating an attack

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "open" and its variants

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

CTV

-0.81

Bey

-0.77

 constitu

-0.71

orum

-0.71

rior

-0.68

opter

-0.63

ensis

-0.62

oria

-0.59

rag

-0.59

POSITIVE LOGITS

 doors

1.12

up

1.04

 Doors

0.96

 portals

0.91

 Pandora

0.88

ource

0.86

 wounds

0.84

 mouths

0.83

 windows

0.82

 fire

0.81

Activations Density 0.058%

the action of physically opening something or initiating an attack

occurrences of the word "open" and its variants

No Comments

No Known Activations