INDEX

Explanations

words related to legal and political matters

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

anwhile

-0.73

 Manhattan

-0.70

 scattering

-0.69

 Send

-0.67

 scatter

-0.67

 exhib

-0.65

 collect

-0.64

Yon

-0.63

 Whit

-0.62

 shack

-0.60

POSITIVE LOGITS

¬

1.33

ľ

1.17

º

1.07

ı

1.04

Ĵ

1.03

Ķ

1.03

¡

1.01

¼

1.01

ĸ

0.99

ĭ

0.99

Activations Density 1.600%

words related to legal and political matters

No Comments

No Known Activations