INDEX

Explanations

special events or significant news-related terms

oai_token-act-pair · gpt-3.5-turbo

references to special news alerts and highlights

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

 alias

-0.68

 Stain

-0.66

 theoret

-0.64

naire

-0.64

 inheritance

-0.61

 shorth

-0.61

ibling

-0.60

ariat

-0.60

 preced

-0.60

urized

-0.59

POSITIVE LOGITS

reddits

0.85

content

0.75

theme

0.70

 content

0.70

 news

0.69

videos

0.67

venture

0.64

Content

0.63

imedia

0.62

axies

0.61

Activations Density 0.093%

special events or significant news-related terms

references to special news alerts and highlights

No Comments

No Known Activations