INDEX

Explanations

elements related to physical appearance and features of a character, such as eye color, facial hair, skin complexion, and body proportions

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

 Anthem

-0.80

rooms

-0.66

acks

-0.62

Â¨

-0.60

vati

-0.59

ivan

-0.59

ifiers

-0.58

mallow

-0.58

 internationally

-0.58

ificent

-0.58

POSITIVE LOGITS

ighty

1.07

gypt

1.02

LECT

0.98

prise

0.95

 plur

0.93

lements

0.87

leanor

0.85

ury

0.85

 coli

0.85

instein

0.84

Activations Density 4.078%

elements related to physical appearance and features of a character, such as eye color, facial hair, skin complexion, and body proportions

No Comments

No Known Activations