INDEX

Explanations

names of people, especially those emphasizing specific letters in their names

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 tremend

-0.91

izoph

-0.70

 DRAG

-0.70

»Ĵ

-0.69

ULAR

-0.68

 glim

-0.68

ikuman

-0.66

ccording

-0.66

 srfAttach

-0.66

eatures

-0.65

POSITIVE LOGITS

kamp

1.04

mann

0.95

berg

0.90

feld

0.88

eman

0.86

wald

0.85

hoff

0.84

berger

0.84

croft

0.83

 Productions

0.82

Activations Density 0.159%

names of people, especially those emphasizing specific letters in their names

No Comments

No Known Activations