INDEX

Explanations

the term "premise" followed by a numeric value indicating the strength of activation

oai_token-act-pair · gpt-3.5-turbo

references to premises in narratives or storylines

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

vor

-0.78

abbling

-0.77

vals

-0.75

cler

-0.73

eder

-0.70

imentary

-0.69

ancers

-0.68

eded

-0.66

itsch

-0.65

iths

-0.64

POSITIVE LOGITS

 premise

1.55

 assumption

0.82

 premises

0.77

 proposition

0.74

REC

0.73

terson

0.68

ulhu

0.67

 underpin

0.67

mot

0.67

fy

0.67

Activations Density 0.007%

the term "premise" followed by a numeric value indicating the strength of activation

references to premises in narratives or storylines

No Comments

No Known Activations