INDEX

Explanations

political and government-related terms and figures

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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

reth

-0.77

units

-0.70

 outputs

-0.69

olutions

-0.68

 reproduction

-0.68

 redress

-0.65

auna

-0.64

 degrees

-0.63

 units

-0.63

olves

-0.63

POSITIVE LOGITS

Jr

1.27

 whom

1.25

who

1.25

Jr

1.22

who

1.20

whose

1.17

 whose

1.13

Sr

1.02

aka

0.98

 Himself

0.82

Activations Density 0.245%

political and government-related terms and figures

No Comments

No Known Activations