INDEX

Explanations

professional titles and names related to specific professions

oai_token-act-pair · gpt-3.5-turbo

proper nouns and titles associated with individuals

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

ILLE

-0.76

caliber

-0.74

zn

-0.72

á¸

-0.71

toggle

-0.70

opez

-0.69

ãĤ¤ãĥĪ

-0.68

ize

-0.68

Ã±

-0.67

Ã±o

-0.65

POSITIVE LOGITS

QC

1.04

 Ukip

1.00

 UKIP

0.97

 lecturer

0.97

BBC

0.96

 criticised

0.96

 prosecuting

0.95

 chairman

0.94

who

0.93

 pictured

0.92

Activations Density 0.121%

professional titles and names related to specific professions

proper nouns and titles associated with individuals

No Comments

No Known Activations