INDEX

Explanations

phrases related to discussions or opinions

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

 Rowe

-0.67

cum

-0.63

personal

-0.59

forms

-0.59

 Rarity

-0.58

 Watt

-0.57

imo

-0.56

REDACTED

-0.54

 Crush

-0.54

ãĤ¯

-0.54

POSITIVE LOGITS

 ourselves

1.40

athered

1.08

bsite

1.01

blogs

1.01

asel

0.99

ibo

0.98

're

0.98

aning

0.96

ird

0.95

IRD

0.94

Activations Density 2.484%

phrases related to discussions or opinions

No Comments

No Known Activations