INDEX

Explanations

dates and locations mentioned in tweets

oai_token-act-pair · gpt-3.5-turbo

dates and specific timestamps associated with events or people

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

itutional

-0.71

zag

-0.66

 periodic

-0.65

 rake

-0.63

 classified

-0.63

umerable

-0.62

netflix

-0.61

 categorized

-0.61

 Forbidden

-0.61

 cultivating

-0.60

POSITIVE LOGITS

 ðŁĳ

0.83

 âĿ

0.82

Dialogue

0.81

âĿ

0.80

EngineDebug

0.76

othy

0.73

âĺ

0.71

Reddit

0.70

Â¯Â¯

0.69

Truth

0.67

Activations Density 0.120%

dates and locations mentioned in tweets

dates and specific timestamps associated with events or people

No Comments

No Known Activations