INDEX

Explanations

phrases indicating seeing, analyzing, or evaluating something in a critical or curious manner

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_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

kus

-0.72

ially

-0.69

haul

-0.65

ilee

-0.65

lear

-0.65

 criticised

-0.64

rus

-0.63

bestos

-0.61

Lear

-0.61

yon

-0.61

POSITIVE LOGITS

why

0.93

 excuses

0.89

 terminating

0.86

enance

0.77

 justifying

0.74

 recommending

0.73

lihood

0.72

 scenario

0.72

 mitigating

0.71

 myself

0.69

Activations Density 0.072%

phrases indicating seeing, analyzing, or evaluating something in a critical or curious manner

No Comments

No Known Activations