INDEX

Explanations

quantities and numerical data

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

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

izont

-0.74

achu

-0.74

aucus

-0.71

hub

-0.68

Hub

-0.68

hare

-0.67

ilater

-0.66

LAN

-0.66

Broad

-0.66

helps

-0.66

POSITIVE LOGITS

 thousand

1.14

 percent

1.09

teen

1.02

 cents

0.97

 hundred

0.94

een

0.94

eenth

0.92

 million

0.84

 pounds

0.82

 hours

0.80

Activations Density 0.023%

quantities and numerical data

No Comments

No Known Activations