INDEX

Explanations

phrases related to giving instructions or advice using the word "let"

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

Zen

-0.67

 Languages

-0.64

availability

-0.61

ombat

-0.58

bage

-0.57

cill

-0.57

 cumbers

-0.56

 holiest

-0.55

cled

-0.55

olars

-0.55

POSITIVE LOGITS

tered

1.23

icia

1.01

tering

1.00

itia

0.92

ting

0.86

 loose

0.79

us

0.72

terness

0.71

 slip

0.70

ugal

0.70

Activations Density 0.373%

phrases related to giving instructions or advice using the word "let"

No Comments

No Known Activations