INDEX

Explanations

questions and phrases about problem-solving and improvement strategies

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

 Hank

-0.17

ertas

-0.16

-0.15

-0.14

 priv

-0.14

uld

-0.14

Fol

-0.14

etwork

-0.13

ully

-0.13

tos

-0.13

POSITIVE LOGITS

fix

0.19

fix

0.19

 remedy

0.19

Fix

0.18

 fixes

0.17

.MixedReality

0.17

 solution

0.17

Fix

0.16

 fixing

0.16

 resolution

0.16

Activations Density 0.082%

questions and phrases about problem-solving and improvement strategies

No Comments

No Known Activations