INDEX

Explanations

phrases involving the government and official entities collaborating or taking action together

oai_token-act-pair · gpt-3.5-turbo

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

ãĤ©

-0.85

ãĤ§

-0.77

ãĤª

-0.73

imens

-0.72

GMT

-0.68

ãĥł

-0.68

å¦

-0.68

otide

-0.67

mol

-0.67

inct

-0.66

POSITIVE LOGITS

 academia

1.25

 judiciary

1.17

 NGOs

1.00

 Congress

0.98

DOJ

0.92

its

0.92

ITS

0.92

 universities

0.91

 regulators

0.91

 nonprofits

0.89

Activations Density 0.231%

phrases involving the government and official entities collaborating or taking action together

No Comments

No Known Activations