INDEX

Explanations

names of people or places

oai_token-act-pair · gpt-3.5-turbo

letters or characters that appear frequently 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

 censored

-0.77

 coli

-0.70

 destro

-0.70

 hairs

-0.70

 envy

-0.70

 kernels

-0.68

 mosqu

-0.68

 proxies

-0.67

 mosaic

-0.67

 prol

-0.67

POSITIVE LOGITS

idd

1.04

afer

0.99

inn

0.99

ady

0.99

urd

0.98

itz

0.98

oor

0.97

acker

0.97

alla

0.96

Ã¼

0.95

Activations Density 0.160%

names of people or places

letters or characters that appear frequently in the text

No Comments

No Known Activations