INDEX

Explanations

mentions of superlatives related to size or ranking

oai_token-act-pair · gpt-3.5-turbo

terms associated with the concept of "largest" or "top" entities within various categories

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 Called

-0.67

 Maryland

-0.66

 Ellison

-0.66

 Kiev

-0.65

USS

-0.64

neys

-0.62

Canada

-0.62

angles

-0.62

 Oregon

-0.61

 Effect

-0.61

POSITIVE LOGITS

 airports

0.77

 diplomat

0.75

 universities

0.74

 citizen

0.73

 democracy

0.73

 ageing

0.71

 electoral

0.70

 newspapers

0.70

 hospitals

0.70

 newspaper

0.70

Activations Density 0.182%

mentions of superlatives related to size or ranking

terms associated with the concept of "largest" or "top" entities within various categories

No Comments

No Known Activations