INDEX

Explanations

phrases related to ordering or sequence

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

tek

-0.86

vae

-0.74

peria

-0.74

attery

-0.73

ky

-0.72

reath

-0.71

tu

-0.71

rities

-0.71

espie

-0.69

Els

-0.69

POSITIVE LOGITS

lies

1.19

liness

0.96

etary

0.90

Osw

0.78

eering

0.73

 Mant

0.68

 fulfillment

0.66

 discipl

0.64

ordering

0.64

 psychiat

0.62

Activations Density 0.614%

phrases related to ordering or sequence

No Comments

No Known Activations