INDEX

Explanations

words related to legal cases and criminal events

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

creen

-0.77

ABE

-0.71

 destro

-0.69

 shroud

-0.67

 Belg

-0.67

 Tasman

-0.67

wagen

-0.67

 Mirage

-0.67

iewicz

-0.66

 Doodle

-0.64

POSITIVE LOGITS

ª

1.31

ł

1.24

Ĳ

1.23

ĳ

1.14

¹

1.08

Ĵ

1.06

ı

1.05

«

1.02

Ķ

0.99

¤

0.98

Activations Density 2.055%

words related to legal cases and criminal events

No Comments

No Known Activations