INDEX

Explanations

connections to false or misleading 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

-,

-0.17

DOMNode

-0.15

orris

-0.14

azz

-0.14

 tern

-0.14

 mxArray

-0.13

.reddit

-0.13

iaux

-0.13

ãĢģäºĮ

-0.13

vault

-0.13

POSITIVE LOGITS

erman

0.16

ombat

0.15

mast

0.15

 fitte

0.14

ÃĹ↵↵

0.14

hic

0.14

alous

0.14

uration

0.14

 .↵↵

0.13

Viewport

0.13

Activations Density 0.045%

connections to false or misleading information

No Comments

No Known Activations