INDEX

Explanations

phrases related to politics and news 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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

é¾įåĸļå£«

-0.64

 foul

-0.61

red

-0.60

 center

-0.58

aways

-0.58

Beg

-0.56

 solicitation

-0.56

 Prompt

-0.55

 vest

-0.55

 centre

-0.55

POSITIVE LOGITS

ichael

1.18

useum

1.16

uppet

1.06

asonic

1.05

onday

1.03

antle

1.03

ixture

1.03

ixed

1.02

ONEY

1.02

igrants

1.02

Activations Density 4.242%

phrases related to politics and news events

No Comments

No Known Activations