INDEX

Explanations

phrases related to nationality, political figures, and ideological concepts

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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

Â·Â·

-0.88

aukee

-0.71

Redditor

-0.67

Courtesy

-0.64

Champ

-0.63

QB

-0.62

iken

-0.61

Tes

-0.60

iste

-0.60

March

-0.60

POSITIVE LOGITS

 subsequent

0.78

 assorted

0.77

 consequ

0.74

 Supplementary

0.67

 consequently

0.66

 occasional

0.64

 countless

0.64

senal

0.63

 partnerships

0.61

 vitro

0.60

Activations Density 0.135%

phrases related to nationality, political figures, and ideological concepts

No Comments

No Known Activations