INDEX

Explanations

phrases related to a numerical comparison, such as "at least"

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

kai

-0.65

axter

-0.64

icides

-0.64

ãĤ¿

-0.63

iday

-0.62

bath

-0.61

ggles

-0.59

tions

-0.59

kefeller

-0.59

gars

-0.59

POSITIVE LOGITS

 partially

0.84

 partly

0.82

 temporarily

0.70

 theoretically

0.68

 SOME

0.68

hap

0.65

foundland

0.64

uner

0.64

lik

0.64

oct

0.62

Activations Density 0.381%

phrases related to a numerical comparison, such as "at least"

No Comments

No Known Activations