INDEX

Explanations

words related to social media usernames or handles, typically with numbers included

oai_token-act-pair · gpt-3.5-turbo

words related to accounting and financial metrics

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 Haku

-0.81

 Butt

-0.76

tob

-0.75

 symp

-0.72

-0.71

 antip

-0.71

INF

-0.70

-0.69

 antit

-0.68

 Ginger

-0.67

POSITIVE LOGITS

ur

1.09

ural

1.05

UR

1.03

ember

0.97

ura

0.97

urrection

0.97

ophon

0.97

ou

0.96

ouch

0.95

urous

0.93

Activations Density 0.299%

words related to social media usernames or handles, typically with numbers included

words related to accounting and financial metrics

No Comments

No Known Activations