INDEX

Explanations

time-related references, such as "years ago," and forum-related timestamps

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

 lasting

-0.22

 governing

-0.22

urring

-0.21

 promise

-0.21

 promises

-0.21

 emerging

-0.19

 underlying

-0.18

 finance

-0.18

 sure

-0.18

 hope

-0.18

POSITIVE LOGITS

emale

0.27

âĹ¼

0.24

 Âł Âł Âł Âł

0.23

tch

0.23

sic

0.23

ensis

0.22

 Reserved

0.21

?????

0.21

zona

0.21

xxxx

0.21

Activations Density 14.540%

time-related references, such as "years ago," and forum-related timestamps

No Comments

No Known Activations