INDEX

Explanations

proper nouns or names ending in 'ras'

oai_token-act-pair · gpt-3.5-turbo

references to a specific term or identifier related to a group's name or individuals

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

 Yankee

-0.71

tails

-0.70

atures

-0.66

tains

-0.65

glers

-0.65

ĵ

-0.65

 Holmes

-0.64

rica

-0.62

 Grimes

-0.58

rican

-0.57

POSITIVE LOGITS

sembly

1.00

vati

0.99

hea

0.97

irez

0.97

xual

0.96

ources

0.96

arin

0.95

bian

0.93

ylum

0.92

hens

0.92

Activations Density 0.017%

proper nouns or names ending in 'ras'

references to a specific term or identifier related to a group's name or individuals

No Comments

No Known Activations