INDEX

Explanations

the number 'one' being used in different contexts across 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_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

ooks

-0.82

ories

-0.74

hips

-0.73

inders

-0.72

osponsors

-0.71

respective

-0.68

actions

-0.63

ourn

-0.63

lations

-0.61

older

-0.60

POSITIVE LOGITS

 hundred

1.21

 Hundred

1.13

 thousand

1.00

 thing

0.93

 dimensional

0.90

 sided

0.88

 Piece

0.83

 Thousand

0.83

 month

0.82

 hour

0.81

Activations Density 0.870%

the number 'one' being used in different contexts across the text

No Comments

No Known Activations