INDEX

Explanations

proper nouns and verbs related to news and events

oai_token-act-pair · gpt-3.5-turbo

content related to disciplinary actions and accusations

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

ntil

-0.61

soDeliveryDate

-0.61

bilt

-0.57

MpServer

-0.55

¶ħ

-0.53

 Slate

-0.52

DragonMagazine

-0.51

ãĤ¨ãĥ«

-0.51

<?

-0.50

quote

-0.50

POSITIVE LOGITS

 arrives

0.62

 unexpectedly

0.61

 enters

0.59

 mysteriously

0.59

 arrived

0.58

 erupted

0.58

 became

0.57

 becomes

0.57

 began

0.56

 came

0.56

Activations Density 1.002%

proper nouns and verbs related to news and events

content related to disciplinary actions and accusations

No Comments

No Known Activations