INDEX

Explanations

names and abbreviations, as well as words related to physical states of being or actions like "sleep," "suicide," and "aggressiveness."

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

Ĥª

-0.66

»Ĵ

-0.63

 tremend

-0.62

ADRA

-0.60

 misdem

-0.58

 bluff

-0.57

 Skydragon

-0.56

 clicks

-0.56

 uninsured

-0.56

 premiums

-0.56

POSITIVE LOGITS

phant

0.81

vre

0.81

ghan

0.75

anasia

0.71

ritch

0.69

rahim

0.69

rencies

0.68

ionage

0.68

naissance

0.67

mot

0.67

Activations Density 4.632%

names and abbreviations, as well as words related to physical states of being or actions like "sleep," "suicide," and "aggressiveness."

No Comments

No Known Activations