INDEX

Explanations

references to legal actions and law enforcement-related activities

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

 Predators

-0.66

berus

-0.61

culosis

-0.61

 Rampage

-0.58

incinn

-0.57

leneck

-0.57

itement

-0.56

ministic

-0.55

Trend

-0.55

issance

-0.54

POSITIVE LOGITS

ridge

1.19

meal

1.06

ridges

0.97

ners

0.90

ials

0.89

icularly

0.89

aking

0.86

icular

0.85

isans

0.83

nered

0.80

Activations Density 3.055%

references to legal actions and law enforcement-related activities

No Comments

No Known Activations