INDEX

Explanations

words related to editing or modifying content, such as "rewritten" or "revised"

oai_token-act-pair · gpt-3.5-turbo

terms related to revising or rewriting content

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

asia

-0.73

stem

-0.66

ccoli

-0.65

Connector

-0.64

ello

-0.63

 overfl

-0.63

¬¼

-0.62

fw

-0.62

places

-0.62

shows

-0.62

POSITIVE LOGITS

é»Ĵ

0.74

 revision

0.73

dated

0.73

 retrospect

0.72

 autobiography

0.71

 chronological

0.70

 guiActiveUn

0.69

 history

0.69

 rewritten

0.67

mbuds

0.66

Activations Density 0.045%

words related to editing or modifying content, such as "rewritten" or "revised"

terms related to revising or rewriting content

No Comments

No Known Activations