INDEX

Explanations

dates and time-related terms

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

EStreamFrame

-0.72

goo

-0.71

ums

-0.64

Condition

-0.63

abilia

-0.63

ogly

-0.60

unta

-0.59

Widget

-0.58

iazep

-0.56

uper

-0.56

POSITIVE LOGITS

we

0.80

 they

0.75

 however

0.72

 there

0.72

 meanwhile

0.71

ruary

0.67

0.63

 rumors

0.62

lehem

0.61

he

0.61

Activations Density 0.773%

dates and time-related terms

No Comments

No Known Activations