INDEX

Explanations

phrases related to news articles, press releases, and journalistic content

oai_token-act-pair · gpt-3.5-turbo

highly impactful medical terms or concepts

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

hement

-0.78

pex

-0.70

 diseng

-0.69

 immobil

-0.66

 clinch

-0.65

thodox

-0.65

hene

-0.65

 destro

-0.64

 decrypt

-0.63

ilers

-0.63

POSITIVE LOGITS

Associated

0.98

News

0.97

Nap

0.96

Topics

0.94

Country

0.89

Narr

0.89

eric

0.84

Temperature

0.83

Department

0.82

Volume

0.81

Activations Density 0.222%

phrases related to news articles, press releases, and journalistic content

highly impactful medical terms or concepts

No Comments

No Known Activations