INDEX

Explanations

the word "look" followed by a number indicating the intensity or manner of looking

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

ä¹

-0.62

Cele

-0.59

orb

-0.58

IL

-0.57

learning

-0.56

MQ

-0.55

ricular

-0.55

 Belief

-0.55

EStreamFrame

-0.55

CENT

-0.55

POSITIVE LOGITS

ahead

1.01

 suspic

0.89

ãĤ¶

0.78

 favorably

0.78

 forward

0.74

 backward

0.74

 alike

0.72

 closely

0.70

iless

0.69

 awfully

0.69

Activations Density 2.503%

the word "look" followed by a number indicating the intensity or manner of looking

No Comments

No Known Activations