INDEX

Explanations

references to authority figures or established standards

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 mathemat

-0.96

 vulner

-0.87

 disadvant

-0.80

 accomp

-0.76

 elig

-0.76

 unnecess

-0.75

 suspic

-0.73

 challeng

-0.72

 mosqu

-0.72

 logger

-0.70

POSITIVE LOGITS

ï¸ı

1.10

own

0.91

ution

0.88

ship

0.86

atural

0.82

sic

0.81

ence

0.81

tal

0.80

iversary

0.80

Activations Density 0.045%

references to authority figures or established standards

No Comments

No Known Activations