INDEX

Explanations

contractions used in negative statements

oai_token-act-pair · gpt-3.5-turbo

negations or phrases expressing something that is not the case

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.11.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

PDATE

-0.88

accompan

-0.83

ñ

-0.77

ò

-0.76

 è£ıè

-0.74

ãĤ¼ãĤ¦ãĤ¹

-0.73

icipated

-0.73

Afee

-0.72

Þ

-0.71

WithNo

-0.70

POSITIVE LOGITS

 necessarily

1.29

 wanna

1.24

 want

1.24

 know

1.17

 need

1.14

 deserve

1.09

 bother

1.08

 seem

1.07

 really

1.06

 hesitate

1.02

Activations Density 0.119%

contractions used in negative statements

negations or phrases expressing something that is not the case

No Comments

No Known Activations