INDEX

Explanations

phrases related to actions or settings

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

ible

-0.76

IBLE

-0.71

ibility

-0.69

ibles

-0.69

ibly

-0.68

issance

-0.67

ory

-0.64

alez

-0.64

ability

-0.64

Sense

-0.63

POSITIVE LOGITS

tle

1.34

 sail

1.02

ters

0.91

abl

0.86

ter

0.84

 forth

0.84

 aside

0.83

itud

0.82

Timeout

0.76

upt

0.75

Activations Density 3.652%

phrases related to actions or settings

No Comments

No Known Activations