INDEX

Explanations

phrases in a specific language or with a unique character pattern

oai_token-act-pair · gpt-3.5-turbo

special characters and non-standard punctuation

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 Stras

-0.76

ellen

-0.70

utterstock

-0.69

ignment

-0.69

ategic

-0.66

ouched

-0.65

ileaks

-0.65

 wart

-0.63

 Wichita

-0.63

warts

-0.62

POSITIVE LOGITS

âĸĵ

1.08

DIT

0.98

ĵ

0.95

¡

0.93

BLE

0.89

Â±

0.87

æµ

0.84

×Ļ

0.84

×ĳ

0.83

uses

0.83

Activations Density 0.005%

phrases in a specific language or with a unique character pattern

special characters and non-standard punctuation

No Comments

No Known Activations