INDEX

Explanations

words related to political locations or events, especially related to Capitol Hill

oai_token-act-pair · gpt-3.5-turbo

references to Capitol Hill

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

cause

-0.73

origin

-0.70

said

-0.68

estamp

-0.67

went

-0.67

nings

-0.66

 Cameroon

-0.65

nir

-0.65

company

-0.64

 Markus

-0.64

POSITIVE LOGITS

 Hill

1.35

 Heights

0.98

itol

0.95

 Square

0.93

Hill

0.92

 Alley

0.89

 dome

0.89

 Theatre

0.87

 Records

0.86

 Capitol

0.86

Activations Density 0.032%

words related to political locations or events, especially related to Capitol Hill

references to Capitol Hill

No Comments

No Known Activations