INDEX

Explanations

phrases related to news events and reactions

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

atl

-0.83

ern

-0.78

soDeliveryDate

-0.78

NET

-0.76

net

-0.74

 isEnabled

-0.73

ise

-0.71

edin

-0.71

wang

-0.71

ieth

-0.69

POSITIVE LOGITS

 daring

1.04

 failing

0.99

 violating

0.96

 refusing

0.93

 having

0.92

geries

0.89

 abandoning

0.88

 breaching

0.88

 lack

0.86

 upholding

0.85

Activations Density 0.682%

phrases related to news events and reactions

No Comments

No Known Activations