INDEX

Explanations

proper nouns and names, potentially related to investigations or scandals

oai_token-act-pair · gpt-3.5-turbo

the letter 't'

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

 thous

-0.77

 conclud

-0.77

[*

-0.71

 Vaugh

-0.71

ãĥ¼ãĥĨ

-0.68

 proport

-0.66

 vulner

-0.65

 ingred

-0.63

ModLoader

-0.62

 detrim

-0.61

POSITIVE LOGITS

zeb

0.68

imo

0.67

mt

0.67

ice

0.64

gallery

0.62

letters

0.61

ush

0.61

info

0.61

amin

0.61

oba

0.61

Activations Density 0.074%

proper nouns and names, potentially related to investigations or scandals

the letter 't'

No Comments

No Known Activations