INDEX

Explanations

phrases related to communication or discussion

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

hews

-0.67

boa

-0.65

aredevil

-0.63

rypt

-0.61

uilt

-0.60

arte

-0.60

~~~~~~~~~~~~~~~~

-0.60

ritional

-0.59

 stocking

-0.59

feeding

-0.59

POSITIVE LOGITS

 about

0.92

 aloud

0.86

 frankly

0.86

 louder

0.85

 ABOUT

0.81

about

0.80

 loudly

0.79

 smack

0.76

 bout

0.75

 candid

0.72

Activations Density 1.288%

phrases related to communication or discussion

No Comments

No Known Activations