INDEX

Explanations

names of authors and years in citations

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

 pride

-0.79

 Confederate

-0.66

 patriotic

-0.65

 surfing

-0.65

 wasteland

-0.64

 vanity

-0.64

 Loaded

-0.64

 cowboy

-0.64

 Destination

-0.63

 Flash

-0.63

POSITIVE LOGITS

et

1.45

owsky

1.17

opoulos

1.10

elli

1.09

itsch

1.06

oglu

1.05

gaard

1.05

asio

1.03

sson

1.02

insky

1.01

Activations Density 0.200%

names of authors and years in citations

No Comments

No Known Activations