INDEX

Explanations

statements where emphasis or stress is put on specific information, possibly related to policy or events

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

DragonMagazine

-0.79

busters

-0.76

PH

-0.69

ä½ľ

-0.68

jan

-0.68

izons

-0.67

Snow

-0.66

kie

-0.66

enium

-0.65

iop

-0.64

POSITIVE LOGITS

 importance

0.96

 emphasis

0.95

 urgency

0.94

 distinctions

0.84

orical

0.83

 emphasize

0.82

 individuality

0.79

 caution

0.78

 emphasized

0.78

iveness

0.78

Activations Density 0.122%

statements where emphasis or stress is put on specific information, possibly related to policy or events

No Comments

No Known Activations