INDEX

Explanations

television-related terms, particularly networks or news organizations

oai_token-act-pair · gpt-3.5-turbo

instances of the acronym "TV" in various contexts related to television

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

leans

-0.68

cific

-0.67

hot

-0.66

0000000000000000

-0.65

ometric

-0.64

tha

-0.64

hammad

-0.63

ibel

-0.62

reci

-0.61

strument

-0.61

POSITIVE LOGITS

ILLE

1.30

 Everywhere

0.92

trop

0.82

NG

0.82

iolet

0.80

PG

0.77

ISION

0.77

ideo

0.77

 Trop

0.76

 Broadcasting

0.75

Activations Density 0.016%

television-related terms, particularly networks or news organizations

instances of the acronym "TV" in various contexts related to television

No Comments

No Known Activations