INDEX

Explanations

words related to strong emotional responses or opinions

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

 Dragonbound

-0.70

 Unch

-0.70

ļé

-0.66

ered

-0.62

lished

-0.61

 Pound

-0.61

erest

-0.60

orld

-0.60

ELS

-0.60

ezvous

-0.59

POSITIVE LOGITS

vana

1.22

ROR

1.08

andom

1.05

rha

1.04

ror

0.94

idium

0.94

bid

0.93

cles

0.91

rored

0.89

mingham

0.87

Activations Density 0.827%

words related to strong emotional responses or opinions

No Comments

No Known Activations