INDEX

Explanations

phrases related to disciplinary actions and procedures

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ileaks

-0.77

ipeg

-0.72

sonian

-0.69

pocket

-0.68

ãĤ¤ãĥĪ

-0.66

izen

-0.66

Hack

-0.65

prop

-0.65

stanbul

-0.64

assies

-0.64

POSITIVE LOGITS

 discipline

0.91

 discipl

0.85

 disciplinary

0.82

HAEL

0.82

 disciplined

0.81

cipline

0.78

 tant

0.78

utical

0.76

iple

0.76

 Discipline

0.73

Activations Density 0.038%

phrases related to disciplinary actions and procedures

No Comments

No Known Activations