INDEX

Explanations

phrases that emphasize remembering or recalling information

oai_token-act-pair · gpt-4o-mini

New Auto-Interp

Configuration

fnlp/Llama-Scope-R1-Distill/800M-Slimpajama-0-OpenR1-Math-220k/L15R

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Hzfinfdu/SlimPajama-3B

Features

32,768

Data Type

float32

Hook Name

blocks.15.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

cerebras/SlimPajama-627B

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ange

-0.18

inde

-0.18

 indeed

-0.17

se

-0.15

 somehow

-0.15

 hack

-0.15

-0.14

ÙĪØ³ÛĮ

-0.14

 suddenly

-0.14

 retiring

-0.14

POSITIVE LOGITS

ibase

0.16

ServiceProvider

0.15

rug

0.15

WithType

0.15

ÏĮÎ³

0.15

Specifier

0.15

tÃ¼

0.14

gorm

0.14

gia

0.14

ToPoint

0.14

Activations Density 0.084%

phrases that emphasize remembering or recalling information

No Comments

No Known Activations