INDEX

Explanations

I'm sorry, I couldn't extract any particular pattern or theme from the activations in neuron 4 for this document

oai_token-act-pair · gpt-3.5-turbo

references to the substance 'esc' or variations of it, likely related to 'escapism'

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

 Surface

-0.71

å¥³

-0.68

 Kinect

-0.63

 Archdemon

-0.63

WP

-0.62

 Monthly

-0.62

 à¨

-0.62

 Lumia

-0.61

 bitch

-0.59

 Avery

-0.59

POSITIVE LOGITS

aped

1.22

esc

1.19

apes

1.16

ribed

1.11

orts

1.11

ript

1.09

autions

1.07

ence

1.07

opes

1.04

apers

1.00

Activations Density 0.005%

I'm sorry, I couldn't extract any particular pattern or theme from the activations in neuron 4 for this document

references to the substance 'esc' or variations of it, likely related to 'escapism'

No Comments

No Known Activations