INDEX

Explanations

strong, emotionally charged language related to political or military discussions

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

['

-0.67

nell

-0.62

 Canaveral

-0.61

æ©

-0.61

"},"

-0.59

 stunts

-0.59

 gull

-0.58

 thriving

-0.58

 valued

-0.58

 thri

-0.57

POSITIVE LOGITS

 summarize

1.20

 recap

1.18

 briefly

1.10

 summar

1.00

 disclaimer

0.97

 spoiler

0.96

Introduction

0.96

endix

0.92

 summarizes

0.89

 reiterate

0.88

Activations Density 3.637%

strong, emotionally charged language related to political or military discussions

No Comments

No Known Activations