INDEX

Explanations

names of events, places, and organizations

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

 unrestricted

-0.65

 secrecy

-0.65

 buzz

-0.63

 stressed

-0.62

 fret

-0.62

 techno

-0.60

 stress

-0.60

 laundering

-0.59

 undecided

-0.59

 Mother

-0.59

POSITIVE LOGITS

ustain

1.22

ixty

1.21

ESSION

1.11

ufficient

1.08

iders

1.06

ometime

1.03

avage

1.02

atellite

1.02

igma

1.00

outher

1.00

Activations Density 3.584%

names of events, places, and organizations

No Comments

No Known Activations