INDEX

Explanations

phrases expressing surprise or disbelief

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

pour

-0.82

ahime

-0.77

 antioxid

-0.73

iant

-0.65

redients

-0.65

åĤ

-0.64

PI

-0.63

achev

-0.63

uliffe

-0.62

imo

-0.61

POSITIVE LOGITS

theless

1.67

 bothered

1.09

 dreamed

1.00

 EVER

0.92

 ceases

0.85

 heard

0.85

 doubted

0.82

 dime

0.81

 bother

0.81

 imagined

0.81

Activations Density 0.292%

phrases expressing surprise or disbelief

No Comments

No Known Activations