INDEX

Explanations

contractions and negated forms

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

itiz

-0.81

 Seasons

-0.66

 Reborn

-0.66

 Opportun

-0.66

çĶŁ

-0.66

å¥

-0.65

ersed

-0.64

 Communities

-0.64

 Rebellion

-0.63

 Citiz

-0.62

POSITIVE LOGITS

 necessarily

1.49

 hesitate

1.17

 bother

1.15

 intend

1.10

 know

1.07

 condone

1.07

 seem

1.06

icably

1.05

 preclude

1.03

 belong

1.03

Activations Density 0.916%

contractions and negated forms

No Comments

No Known Activations