INDEX

Explanations

phrases related to ensuring or confirming something

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

•"make sure"

"make sure"

•Make sure feature

•Canonical Figure 13

No Comments

Negative Logits

cffffcc

-0.82

pmwiki

-0.76

 ç¥ŀ

-0.71

insula

-0.71

EStreamFrame

-0.69

gression

-0.67

CSS

-0.67

zie

-0.66

 Cosponsors

-0.65

question

-0.64

POSITIVE LOGITS

 nobody

0.73

 they

0.67

ariat

0.67

 everything

0.67

teness

0.65

ially

0.65

Uz

0.65

rity

0.65

worthiness

0.64

recy

0.62

Activations Density 1.744%

phrases related to ensuring or confirming something

No Comments

No Known Activations