INDEX

Explanations

phrases expressing relief and positive outcomes

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ãĥ¯ãĥ³

-0.69

 interfering

-0.67

 improper

-0.66

apons

-0.65

viol

-0.65

umes

-0.64

 prohibited

-0.62

hid

-0.62

 invasive

-0.61

 obstruct

-0.60

POSITIVE LOGITS

 reassure

0.93

 reassured

0.89

 welcome

0.86

 thanking

0.83

 reassuring

0.83

 assurance

0.82

 upbeat

0.81

hearted

0.78

 relief

0.78

 congratulations

0.78

Activations Density 0.327%

phrases expressing relief and positive outcomes

No Comments

No Known Activations