INDEX

Explanations

mentions of the word "presidential"

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.8.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

bane

-0.90

âķĲ

-0.78

bars

-0.78

galitarian

-0.78

âķĲâķĲ

-0.75

MQ

-0.75

DL

-0.75

ener

-0.74

ÙĴ

-0.71

 à¨

-0.71

POSITIVE LOGITS

 election

1.31

 candidate

1.21

 hopeful

1.19

 elections

1.18

 contender

1.12

 candidates

1.12

 nominee

1.10

 administrations

1.07

 aspir

1.06

 nominating

1.05

Activations Density 0.025%

mentions of the word "presidential"

No Comments

No Known Activations