INDEX

Explanations

phrases and terms related to various forms of misconduct or unethical behavior

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

stanbul

-0.91

izen

-0.88

iku

-0.78

osphere

-0.75

pop

-0.73

IPS

-0.68

UNCH

-0.67

ften

-0.67

mega

-0.66

etically

-0.66

POSITIVE LOGITS

 misconduct

1.15

 allegations

1.05

onduct

0.95

 perpetrated

0.88

 scandals

0.88

 accusations

0.86

 wrongdoing

0.86

 disclosures

0.84

 malf

0.83

 manslaughter

0.81

Activations Density 0.007%

phrases and terms related to various forms of misconduct or unethical behavior

No Comments

No Known Activations