INDEX

Explanations

initially, proper nouns, names, or other unique identifiers

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

ngth

-0.63

pter

-0.63

onym

-0.60

 Sorce

-0.57

 Remastered

-0.56

igraph

-0.55

 Torment

-0.55

 Xperia

-0.55

Nex

-0.55

 Vega

-0.55

POSITIVE LOGITS

levard

0.98

apest

0.97

lehem

0.91

ĵĺ

0.87

aneers

0.86

abase

0.78

illet

0.78

reau

0.77

mingham

0.74

pillar

0.74

Activations Density 1.398%

initially, proper nouns, names, or other unique identifiers

No Comments

No Known Activations