INDEX

Explanations

references to the country "Canada."

oai_token-act-pair · gpt-3.5-turbo

the word "Can" in various contexts throughout the document

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

 striving

-0.67

é¾įå¥ĳå£«

-0.64

 Confeder

-0.60

 dishon

-0.59

BAL

-0.59

 rehears

-0.58

 honoring

-0.58

 reven

-0.58

 aven

-0.56

 rehearsal

-0.55

POSITIVE LOGITS

't

1.44

berra

1.36

adian

1.33

vas

1.32

isters

1.13

opy

1.11

ister

1.07

nery

1.03

icum

0.95

ibus

0.95

Activations Density 0.057%

references to the country "Canada."

the word "Can" in various contexts throughout the document

No Comments

No Known Activations