INDEX

Explanations

phrases and words indicating an evaluation or judgment, typically with a negative connotation

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

prus

-0.73

tsky

-0.70

rote

-0.70

utm

-0.70

iHUD

-0.69

gypt

-0.68

rain

-0.67

release

-0.66

ortment

-0.65

thur

-0.65

POSITIVE LOGITS

 anymore

1.01

 enough

0.96

nor

0.96

 whatsoever

0.89

 Enough

0.78

enough

0.76

 anywhere

0.72

 either

0.69

 anybody

0.68

 except

0.67

Activations Density 0.082%

phrases and words indicating an evaluation or judgment, typically with a negative connotation

No Comments

No Known Activations