INDEX

Explanations

political and governmental references, including discussions, controversies, and meetings

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

inction

-0.79

clips

-0.75

wcsstore

-0.69

mol

-0.68

yt

-0.68

verage

-0.67

inventoryQuantity

-0.67

ense

-0.66

urgy

-0.66

ciation

-0.65

POSITIVE LOGITS

 fellow

1.35

 representatives

1.29

 stakeholders

1.22

 superiors

1.21

 colleagues

1.19

 businessmen

1.17

 individuals

1.16

 locals

1.15

 passers

1.14

 executives

1.14

Activations Density 7.017%

political and governmental references, including discussions, controversies, and meetings

No Comments

No Known Activations