INDEX

Explanations

initials in a specific format - T.R.I.P.E., along with specific structured titles including A.L.A. and Jr

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

theless

-0.73

schild

-0.65

adobe

-0.64

 evenings

-0.63

 favour

-0.59

 discour

-0.57

 caution

-0.55

 caps

-0.55

 centre

-0.55

 whisk

-0.55

POSITIVE LOGITS

.,

1.99

.?

1.71

.:

1.52

.;

1.51

.,"

1.50

./

1.31

.—

1.23

.),

1.21

.-

1.14

orporated

1.14

Activations Density 1.418%

initials in a specific format - T.R.I.P.E., along with specific structured titles including A.L.A. and Jr

No Comments

No Known Activations