INDEX

Explanations

statements or arguments effectively disagreeing with a particular viewpoint or concept in the text

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

chenko

-0.68

marine

-0.66

zona

-0.66

è¦ļéĨĴ

-0.63

Eat

-0.59

olan

-0.54

smanship

-0.54

urat

-0.54

hao

-0.54

 forg

-0.53

POSITIVE LOGITS

bidden

1.51

 example

1.51

 instance

1.47

cing

1.38

gotten

1.36

ced

1.24

 starters

1.21

give

1.19

getting

1.11

cible

1.05

Activations Density 0.322%

statements or arguments effectively disagreeing with a particular viewpoint or concept in the text

No Comments

No Known Activations