INDEX

Explanations

terms related to breaking or damaging objects in a live performance or event

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_fs12288-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

12,288

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

fur

-0.72

grad

-0.71

eli

-0.70

win

-0.70

rock

-0.69

mot

-0.66

hon

-0.66

Publisher

-0.65

habi

-0.65

beat

-0.64

POSITIVE LOGITS

 theirs

1.33

 hers

1.12

 them

1.01

 something

1.00

 anything

1.00

 yours

0.97

 another

0.91

 everything

0.90

 ours

0.86

 THEM

0.82

Activations Density 0.828%

terms related to breaking or damaging objects in a live performance or event

No Comments

No Known Activations