INDEX

Explanations

terms related to abuse and abusive behavior

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

cil

-0.80

travel

-0.79

pard

-0.76

izen

-0.75

views

-0.73

ript

-0.69

cise

-0.68

pai

-0.67

vision

-0.65

 Solitaire

-0.64

POSITIVE LOGITS

abuse

0.96

 perpetrated

0.93

 abusing

0.90

 inflicted

0.90

 abuse

0.88

 victims

0.82

 abused

0.79

 abuses

0.79

 abusers

0.79

 survivors

0.73

Activations Density 0.024%

terms related to abuse and abusive behavior

No Comments

No Known Activations