INDEX

Explanations

country names and plane-related terms

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

 Salvation

-0.71

ALK

-0.64

 Polo

-0.64

ITED

-0.64

 Carbuncle

-0.62

 Leap

-0.62

 NETWORK

-0.61

 Shutterstock

-0.60

ente

-0.59

YA

-0.58

POSITIVE LOGITS

colm

1.23

practice

1.11

ibu

0.99

adies

0.98

formed

0.91

icious

0.88

querade

0.88

oney

0.83

aques

0.82

nown

0.82

Activations Density 0.720%

country names and plane-related terms

No Comments

No Known Activations