INDEX

Explanations

references to subjects and topics being discussed or reported on

oai_token-act-pair · gpt-4o-mini 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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

respect

-0.67

iolet

-0.67

ipolar

-0.61

ategor

-0.56

atomic

-0.56

 paran

-0.55

represent

-0.55

apons

-0.54

hem

-0.54

oho

-0.53

POSITIVE LOGITS

liest

1.33

iest

1.17

 thereof

0.96

of

0.86

 centerpiece

0.80

ultimate

0.78

waters

0.73

most

0.72

hest

0.72

 cornerstone

0.68

Activations Density 0.098%

references to subjects and topics being discussed or reported on

No Comments

No Known Activations