INDEX

Explanations

proper nouns such as locations, names, and organizations

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

 congr

-0.67

 deduct

-0.60

 cabinets

-0.60

lc

-0.59

ichick

-0.59

 beetles

-0.56

course

-0.56

rog

-0.55

 additionally

-0.55

 immobil

-0.55

POSITIVE LOGITS

WASHINGTON

1.11

SAN

0.94

ccording

0.90

LOS

0.89

everal

0.88

SHARE

0.85

Associated

0.84

Washington

0.84

TOR

0.83

Madison

0.82

Activations Density 2.821%

proper nouns such as locations, names, and organizations

No Comments

No Known Activations