INDEX

Explanations

phrases related to tasks, procedures, or steps

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

blogspot

-0.65

 incidentally

-0.64

never

-0.62

quel

-0.62

 bragging

-0.62

pmwiki

-0.62

always

-0.60

 Commando

-0.60

 never

-0.58

 seldom

-0.58

POSITIVE LOGITS

enough

0.90

 properly

0.76

pleted

0.71

 fully

0.70

iphany

0.69

 satisfactory

0.68

uate

0.67

 fruition

0.67

 satisf

0.65

 oneself

0.65

Activations Density 0.246%

phrases related to tasks, procedures, or steps

No Comments

No Known Activations