INDEX

Explanations

references to humor-related words or phrases

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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

hips

-0.80

 Dele

-0.69

 Borough

-0.64

 DARK

-0.63

 harbour

-0.63

cape

-0.63

 Leviathan

-0.62

 approve

-0.59

 EntityItem

-0.59

 Construct

-0.58

POSITIVE LOGITS

pty

1.32

orously

1.17

mus

1.12

ulus

1.11

mers

1.10

iliation

1.08

ility

1.08

ankind

1.05

ming

1.03

iday

1.01

Activations Density 0.025%

references to humor-related words or phrases

No Comments

No Known Activations