INDEX

Explanations

names of individuals or characters

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 limits

-0.64

telling

-0.63

\/\/

-0.58

ngth

-0.58

 doom

-0.57

 corners

-0.56

imposed

-0.56

pert

-0.56

glers

-0.54

 spiders

-0.53

POSITIVE LOGITS

lehem

0.76

ÃĽ

0.76

abase

0.73

asion

0.73

ĪĴ

0.71

ainment

0.68

uration

0.64

asia

0.63

itably

0.62

aez

0.62

Activations Density 4.648%

names of individuals or characters

No Comments

No Known Activations