INDEX

Explanations

phrases indicating the extreme or peak level of something

oai_token-act-pair · gpt-3.5-turbo

references to extremes or superlatives in a context

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

FTWARE

-0.83

imaru

-0.75

bill

-0.71

kamp

-0.66

 Tayyip

-0.64

eria

-0.62

dit

-0.61

ouple

-0.61

vich

-0.60

ela

-0.60

POSITIVE LOGITS

 disposal

1.08

 weakest

1.06

 inception

1.05

 heaviest

1.04

 fingertips

1.00

 peak

1.00

 fullest

0.99

 lowest

0.96

 busiest

0.95

 worst

0.94

Activations Density 0.031%

phrases indicating the extreme or peak level of something

references to extremes or superlatives in a context

No Comments

No Known Activations