INDEX

Explanations

people's names, potentially related to crimes or legal matters

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

 Fract

-0.75

BIP

-0.74

 Gemini

-0.72

âĵĺ

-0.72

MODE

-0.72

ãĥĩ

-0.69

!/

-0.67

ãĥ¼ãĤ¯

-0.67

 Galileo

-0.65

 Tibetan

-0.65

POSITIVE LOGITS

aughlin

1.26

endon

0.98

erm

0.86

arks

0.85

ussen

0.84

enn

0.84

uggets

0.83

ough

0.83

atch

0.82

arty

0.81

Activations Density 0.015%

people's names, potentially related to crimes or legal matters

No Comments

No Known Activations