INDEX

Explanations

terms related to the United States (U.S.)

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 rumours

-0.63

 Ernst

-0.63

 volumes

-0.62

 Brill

-0.62

 partName

-0.62

 organisers

-0.62

 Franch

-0.61

 tracts

-0.61

HF

-0.60

 Guinness

-0.58

POSITIVE LOGITS

 Constitution

0.95

 Sovere

0.94

ctions

0.79

military

0.78

 sovereignty

0.74

 democracy

0.74

 economy

0.73

citizens

0.73

security

0.72

ensable

0.71

Activations Density 0.138%

terms related to the United States (U.S.)

No Comments

No Known Activations