INDEX

Explanations

names of individuals and references to political events

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

 ç¥ŀ

-1.13

enment

-0.90

}}}

-0.86

iku

-0.85

liga

-0.85

 rejoice

-0.83

uku

-0.81

ciation

-0.80

liam

-0.80

 âĿ

-0.80

POSITIVE LOGITS

 alleged

1.43

 incrim

1.41

 confidential

1.34

 informant

1.33

 allegedly

1.32

 informants

1.28

 leaked

1.27

 dossier

1.27

 wiret

1.27

 whistleblower

1.25

Activations Density 6.264%

names of individuals and references to political events

No Comments

No Known Activations