INDEX

Explanations

contradictory statements or stances

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

odan

-0.70

 anonymity

-0.66

geons

-0.64

missions

-0.64

gee

-0.64

 Examiner

-0.62

iger

-0.61

tor

-0.61

asio

-0.60

 audience

-0.59

POSITIVE LOGITS

 else

1.55

Else

1.13

 resembling

1.04

 Else

1.00

 imaginable

0.92

 pertaining

0.88

 happening

0.87

 happens

0.86

NESS

0.81

else

0.81

Activations Density 0.341%

contradictory statements or stances

No Comments

No Known Activations