INDEX

Explanations

references to educational institutions or related phrases

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

•Linear education level

No Comments

Negative Logits

leans

-0.70

 Hive

-0.67

ioxide

-0.66

 heel

-0.61

neum

-0.59

OPLE

-0.59

 footprints

-0.58

 Bundy

-0.58

obin

-0.58

 genus

-0.58

POSITIVE LOGITS

girls

1.44

children

1.37

kids

1.25

girl

1.21

boys

1.08

 curriculum

1.06

boy

1.05

masters

1.04

master

1.03

 Teachers

0.99

Activations Density 0.411%

references to educational institutions or related phrases

No Comments

No Known Activations