INDEX

Explanations

phrases or concepts related to attention or focus

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ihad

-0.85

thia

-0.68

fortune

-0.68

aves

-0.67

tein

-0.67

assian

-0.66

stead

-0.65

cible

-0.64

went

-0.63

tarian

-0.61

POSITIVE LOGITS

 spotlight

1.06

 shone

1.04

elight

1.02

 Spotlight

0.92

 shine

0.91

 shines

0.88

eless

0.86

 shining

0.77

 prominently

0.73

 notch

0.72

Activations Density 0.028%

phrases or concepts related to attention or focus

No Comments

No Known Activations