INDEX

Explanations

phrases related to actions or concepts being moved away from

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

umn

-0.70

ums

-0.69

oly

-0.68

ummer

-0.66

 Briggs

-0.65

xual

-0.65

ipel

-0.64

ionic

-0.64

milo

-0.64

iann

-0.62

POSITIVE LOGITS

 from

0.78

fitted

0.73

away

0.72

 FROM

0.70

finder

0.70

ired

0.70

ï¸

0.69

heid

0.68

ruck

0.68

fits

0.67

Activations Density 0.505%

phrases related to actions or concepts being moved away from

No Comments

No Known Activations