INDEX

Explanations

terms related to reality

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

xual

-0.83

theless

-0.73

 Vaugh

-0.65

ucky

-0.65

 Carbuncle

-0.61

indal

-0.60

served

-0.60

anted

-0.59

VC

-0.58

 clipboard

-0.58

POSITIVE LOGITS

ignment

1.32

polit

1.20

estate

1.17

isation

1.11

 estate

1.11

igning

1.01

izations

0.98

igned

0.96

izable

0.93

DonaldTrump

0.90

Activations Density 3.420%

terms related to reality

No Comments

No Known Activations