INDEX

Explanations

words related to proper names or entities, potentially including surnames or initials

oai_token-act-pair · gpt-3.5-turbo

initials or abbreviations in names or titles

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

xp

-0.73

 SHARES

-0.63

 fools

-0.62

plays

-0.62

 subscript

-0.61

Fn

-0.61

 SPACE

-0.58

 profit

-0.58

wordpress

-0.58

 capitals

-0.58

POSITIVE LOGITS

Hol

0.83

rimination

0.74

HR

0.73

rim

0.70

az

0.69

oldemort

0.68

illard

0.68

adan

0.67

 Stores

0.67

shut

0.66

Activations Density 0.139%

words related to proper names or entities, potentially including surnames or initials

initials or abbreviations in names or titles

No Comments

No Known Activations