INDEX

Explanations

terms related to politics and public figures

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

igslist

-0.75

ãĤ¶

-0.74

rote

-0.73

 Previously

-0.72

eln

-0.71

ĵ

-0.70

bis

-0.70

Hide

-0.69

ror

-0.69

Ö¼

-0.68

POSITIVE LOGITS

 whichever

1.03

 whatever

0.89

 regardless

0.86

 ALWAYS

0.85

whatever

0.81

 wherever

0.80

 whether

0.79

 depending

0.79

 invariably

0.77

 irrespective

0.77

Activations Density 0.246%

terms related to politics and public figures

No Comments

No Known Activations