INDEX

Explanations

proper nouns related to geographical locations and titles of authority figures

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

iating

-0.72

chell

-0.70

mble

-0.70

orer

-0.68

ters

-0.68

idently

-0.68

ombies

-0.67

TING

-0.67

OUT

-0.67

heric

-0.65

POSITIVE LOGITS

pin

1.18

uin

1.14

doms

1.14

pins

1.05

dom

1.02

DOM

0.99

 Arabian

0.93

 Arabia

0.92

 Abdullah

0.84

fish

0.80

Activations Density 0.960%

proper nouns related to geographical locations and titles of authority figures

No Comments

No Known Activations