INDEX

Explanations

words related to software or technology releases

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

hov

-0.70

isma

-0.68

gling

-0.67

esthetic

-0.64

atomic

-0.64

fine

-0.63

mong

-0.62

orum

-0.62

pn

-0.60

neg

-0.60

POSITIVE LOGITS

 announcements

0.95

 RELE

0.79

tion

0.77

new

0.75

announced

0.75

tions

0.73

dates

0.72

date

0.71

 notices

0.71

éĹĺ

0.71

Activations Density 5.020%

words related to software or technology releases

No Comments

No Known Activations