INDEX

Explanations

names, especially those related to family lineage and heritage

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

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

safe

-0.76

asive

-0.76

 airspace

-0.75

 safer

-0.73

mbuds

-0.73

Snake

-0.70

 usability

-0.70

cedented

-0.69

 safe

-0.68

CONCLUS

-0.67

POSITIVE LOGITS

aceae

0.93

lda

0.92

 paternal

0.87

idae

0.86

Sr

0.83

*)

0.83

 Rodham

0.82

 grandfather

0.82

Tud

0.79

 Saud

0.79

Activations Density 0.135%

names, especially those related to family lineage and heritage

No Comments

No Known Activations