INDEX

Explanations

Roman numerals representing historical figures or events

oai_token-act-pair · gpt-3.5-turbo

Roman numerals associated with historical figures or events

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

•GPT2 L5 Resid: RRLRRRRRRRRRRLRRRLLRRRRRRLL

No Comments

Negative Logits

bies

-0.88

boards

-0.87

feed

-0.81

href

-0.80

sight

-0.78

bery

-0.75

lins

-0.73

ãĥ¡

-0.71

cho

-0.70

ãĤ±

-0.69

POSITIVE LOGITS

III

1.21

II

0.97

 Jinping

0.88

III

0.79

 VIII

0.77

oodoo

0.74

arios

0.74

Aus

0.71

 Centauri

0.70

VII

0.69

Activations Density 0.017%

Roman numerals representing historical figures or events

Roman numerals associated with historical figures or events

No Comments

No Known Activations