INDEX

Explanations

mentions of rankings or ordering based on performance or categories

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_fs12288-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

12,288

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 disse

-0.77

 Emails

-0.75

soType

-0.73

catentry

-0.72

 overflow

-0.71

 gestures

-0.71

illance

-0.71

ewitness

-0.71

 informants

-0.71

 shielding

-0.71

POSITIVE LOGITS

 Trinidad

0.96

 Finland

0.95

 Kansas

0.95

 Georgia

0.95

 lowly

0.94

 Saint

0.93

 Norway

0.93

 Switzerland

0.90

 Alabama

0.88

 Minnesota

0.88

Activations Density 0.803%

mentions of rankings or ordering based on performance or categories

No Comments

No Known Activations