INDEX

Explanations

things not happening or being negated in the text

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

.","

-0.79

itiz

-0.73

-"

-0.71

 thereto

-0.70

 ..."

-0.65

—"

-0.63

``(

-0.63

 ÂŃ

-0.62

Ïī

-0.61

[|

-0.60

POSITIVE LOGITS

resa

1.59

odore

1.26

xiety

1.20

alyst

1.04

mosp

1.03

withstanding

1.00

swers

1.00

bidden

0.94

anmar

0.91

chieve

0.90

Activations Density 3.124%

things not happening or being negated in the text

No Comments

No Known Activations