INDEX

Explanations

social media or communication-related headings

oai_token-act-pair · gpt-3.5-turbo

occurrences of the character 'w/' in the text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

agos

-0.72

 obliged

-0.65

 Reconstruction

-0.65

emort

-0.63

yrinth

-0.62

 snipp

-0.62

 Lumpur

-0.60

istar

-0.59

vertis

-0.59

terness

-0.59

POSITIVE LOGITS

sole

0.68

stood

0.66

hold

0.64

AFP

0.64

lain

0.64

lde

0.63

Mot

0.63

ÏĢ

0.62

ste

0.61

ATIONS

0.61

Activations Density 0.025%

social media or communication-related headings

occurrences of the character 'w/' in the text

No Comments

No Known Activations