INDEX

Explanations

phrases related to contrasting or opposing concepts

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

auga

-0.74

cised

-0.66

wings

-0.63

 Frie

-0.63

ardless

-0.62

mentation

-0.61

 Regions

-0.61

enhagen

-0.61

lighting

-0.61

gio

-0.60

POSITIVE LOGITS

 innocuous

1.02

 daunting

0.97

 intuitive

0.93

 clichÃ©

0.91

oxy

0.91

intuitive

0.86

 confusing

0.85

 contradictory

0.83

 contradiction

0.83

 intimidating

0.82

Activations Density 0.221%

phrases related to contrasting or opposing concepts

No Comments

No Known Activations