INDEX

Explanations

the string "rah" with varying activations

oai_token-act-pair · gpt-3.5-turbo

proper nouns, particularly names related to the subject matter

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Rah

-1.36

ÄŁ

-0.95

Erd

-0.84

oÄŁ

-0.74

QR

-0.67

 Iter

-0.67

Rah

-0.66

 Uber

-0.66

 Ankara

-0.66

ÅŁ

-0.66

POSITIVE LOGITS

rah

2.85

grid

1.46

 Damon

1.31

 Donovan

1.07

aline

1.03

HF

1.00

lene

1.00

orah

0.96

ilda

0.93

rina

0.92

Activations Density 0.025%

the string "rah" with varying activations

proper nouns, particularly names related to the subject matter

No Comments

No Known Activations