INDEX

Explanations

words related to news reporting and events

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

ãĥ¯ãĥ³

-0.83

hei

-0.75

ãĥķãĤ©

-0.71

æ³

-0.69

 Helic

-0.69

é¾įå

-0.68

ãĤ®

-0.68

è¡

-0.67

 Saras

-0.67

æĸ

-0.66

POSITIVE LOGITS

vious

0.71

]);

0.70

ctuary

0.66

ption

0.63

eters

0.63

pless

0.62

kered

0.62

ument

0.62

¼

0.60

aming

0.60

Activations Density 1.275%

words related to news reporting and events

No Comments

No Known Activations