INDEX

Explanations

contractions involving "don't"

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

itiz

-0.71

 Reborn

-0.71

soType

-0.70

ersed

-0.69

 Laun

-0.63

çĶŁ

-0.63

 Communities

-0.63

 Rebellion

-0.61

 Opportun

-0.60

Reviewer

-0.60

POSITIVE LOGITS

 necessarily

1.47

 bother

1.27

 hesitate

1.23

 condone

1.07

 discriminate

1.06

 intend

1.06

 know

1.06

 seem

1.03

 specify

1.02

 preclude

1.01

Activations Density 1.774%

contractions involving "don't"

No Comments

No Known Activations