INDEX

Explanations

instances of information being intentionally kept from individuals or the public

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

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

psc

-0.85

isol

-0.77

gdala

-0.77

plex

-0.75

onies

-0.71

liam

-0.71

swick

-0.69

scape

-0.69

creation

-0.68

hran

-0.67

POSITIVE LOGITS

 gratification

0.90

 withheld

0.83

 withholding

0.81

 withhold

0.81

otal

0.80

 confidential

0.80

 information

0.79

 refunds

0.77

 payment

0.76

erential

0.73

Activations Density 0.009%

instances of information being intentionally kept from individuals or the public

No Comments

No Known Activations