INDEX

Explanations

names of U.S. states

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

BOOK

-0.73

fen

-0.73

henko

-0.70

byn

-0.70

 Britann

-0.70

deck

-0.69

colo

-0.68

htar

-0.67

ynt

-0.67

lihood

-0.66

POSITIVE LOGITS

 State

1.19

 Springs

1.17

Gov

1.03

 Governor

0.99

 Tech

0.98

 Rapids

0.98

sylvania

0.96

 Legislature

0.94

 Attorney

0.91

 Railroad

0.90

Activations Density 1.129%

names of U.S. states

No Comments

No Known Activations