INDEX

Explanations

proper nouns or place names with diacritics

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

 multiplication

-0.66

 logger

-0.65

 agre

-0.65

 masturbation

-0.64

 patched

-0.63

 checkout

-0.62

geries

-0.61

 kernels

-0.61

 unborn

-0.60

 clipping

-0.60

POSITIVE LOGITS

ï¸ı

1.28

ÃĽ

0.97

Â£

0.96

Ã±

0.92

¸

0.91

Ëľ

0.91

ÃįÃį

0.91

ĺ

0.90

ĳ

0.90

Â½

0.89

Activations Density 0.057%

proper nouns or place names with diacritics

No Comments

No Known Activations