INDEX

Explanations

items related to technical specifications, possibly in a categorized list format

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

ĳå£«

-0.68

retty

-0.67

 certific

-0.66

 allowances

-0.63

 subur

-0.63

 practition

-0.62

ģĸ

-0.60

 unbeliev

-0.60

icho

-0.60

ataka

-0.59

POSITIVE LOGITS

theless

1.05

bor

0.77

stadt

0.75

haus

0.75

west

0.74

ukong

0.70

armac

0.69

phrine

0.69

bors

0.68

bour

0.67

Activations Density 0.046%

items related to technical specifications, possibly in a categorized list format

No Comments

No Known Activations