INDEX

Explanations

phrases that indicate caution or hesitation before reaching a conclusion

oai_token-act-pair · gpt-4o-mini

New Auto-Interp

Configuration

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

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

open-r1/OpenR1-Math-220k

Features

32,768

Data Type

float32

Hook Name

blocks.15.hook_resid_post

Architecture

jumprelu

Context Size

1,024

Dataset

cerebras/SlimPajama-627B

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

rey

-0.07

atches

-0.07

adt

-0.06

ary

-0.06

pa

-0.06

 inline

-0.06

 intr

-0.06

stry

-0.06

ener

-0.06

aur

-0.06

POSITIVE LOGITS

 anything

0.08

anything

0.06

ncoder

0.06

DBus

0.06

ANGO

0.06

GOR

0.06

lue

0.06

urn

0.06

à¹Īà¸²à¸ĩà¸ģ

0.06

 manip

0.06

Activations Density 0.046%

phrases that indicate caution or hesitation before reaching a conclusion

No Comments

No Known Activations