INDEX

Explanations

legal or criminal terms related to prohibited activities

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

 gigs

-0.84

 Roose

-0.82

 moms

-0.77

 booze

-0.77

 Philly

-0.75

 crunch

-0.74

 Tough

-0.73

 veggies

-0.73

Crystal

-0.72

 debunk

-0.72

POSITIVE LOGITS

 pursuant

1.42

 hereby

1.41

 lawful

1.32

 lawfully

1.29

 practicable

1.24

 specified

1.22

 unlawfully

1.19

 persons

1.19

 applicable

1.18

 unlawful

1.17

Activations Density 3.211%

legal or criminal terms related to prohibited activities

No Comments

No Known Activations