INDEX

Explanations

phrases related to news articles and information reporting on events

oai_token-act-pair · gpt-3.5-turbo

punctuation and format indicators in 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

 endeavour

-0.63

 Emin

-0.63

iott

-0.63

 domain

-0.61

Dyn

-0.60

cess

-0.60

 overtake

-0.60

 Squadron

-0.58

 reiter

-0.58

 hect

-0.57

POSITIVE LOGITS

LOS

1.01

WASHINGTON

0.96

SAN

0.96

ccording

0.85

NEW

0.84

Born

0.80

Copyright

0.77

SHARE

0.77

HEAD

0.76

BAT

0.75

Activations Density 0.104%

phrases related to news articles and information reporting on events

punctuation and format indicators in text

No Comments

No Known Activations