INDEX

Explanations

phrases indicating levels of comparison or preference

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

 understandably

-0.61

rocket

-0.60

orage

-0.60

ords

-0.59

 pores

-0.59

VERTISEMENT

-0.58

restling

-0.55

ERSON

-0.55

 therefore

-0.53

 Jakarta

-0.53

POSITIVE LOGITS

abouts

1.24

Else

1.16

phans

1.01

 alternatively

0.95

nam

0.94

depending

0.92

 equivalent

0.89

acle

0.89

chard

0.85

 equival

0.85

Activations Density 0.226%

phrases indicating levels of comparison or preference

No Comments

No Known Activations