INDEX

Explanations

instances of a specific term "endor" mentioned in various contexts

oai_token-act-pair · gpt-3.5-turbo

references to endorsements or supporting statements

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 gravy

-0.71

oola

-0.69

 reps

-0.66

 hypers

-0.65

 Teach

-0.64

alph

-0.64

Austral

-0.63

 recons

-0.62

ographically

-0.62

CHO

-0.62

POSITIVE LOGITS

endor

3.85

governmental

1.34

enstein

1.23

DW

1.18

ember

1.05

0.98

DEN

0.91

linger

0.90

Contribut

0.89

eking

0.89

Activations Density 0.045%

instances of a specific term "endor" mentioned in various contexts

references to endorsements or supporting statements

No Comments

No Known Activations