INDEX

Explanations

words related to negative attributes and actions, such as incompetence, stupidity, and thoughtlessness

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

amins

-0.81

nesium

-0.74

ergy

-0.72

amen

-0.70

ãĥİ

-0.67

lift

-0.65

izable

-0.64

ellar

-0.60

lyak

-0.60

ITNESS

-0.59

POSITIVE LOGITS

 perv

1.06

iness

1.02

 inherent

0.95

fulness

0.88

ously

0.84

 quot

0.84

 bordering

0.82

 abound

0.80

ulence

0.80

 crept

0.79

Activations Density 0.143%

words related to negative attributes and actions, such as incompetence, stupidity, and thoughtlessness

No Comments

No Known Activations