INDEX

Explanations

imperative verbs followed by direct objects

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

 Languages

-0.65

Zen

-0.65

availability

-0.63

Interstitial

-0.63

inger

-0.62

bage

-0.62

liest

-0.60

marine

-0.59

DAQ

-0.58

natureconservancy

-0.58

POSITIVE LOGITS

tered

1.14

tering

0.95

icia

0.95

us

0.92

me

0.80

 loose

0.78

itia

0.78

him

0.76

ting

0.75

 slip

0.73

Activations Density 2.857%

imperative verbs followed by direct objects

No Comments

No Known Activations