INDEX

Explanations

expressions of correctness versus incorrectness in arguments

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

New Auto-Interp

Configuration

fnlp/Llama3_1-8B-Base-LXM-8x/Llama3_1-8B-Base-L22M-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.22.hook_mlp_out

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

 correct

-0.23

 right

-0.20

correct

-0.19

-right

-0.18

Correct

-0.17

_correct

-0.17

ined

-0.17

 serialVersionUID

-0.15

ambi

-0.15

 Correct

-0.15

POSITIVE LOGITS

roup

0.18

jerne

0.17

ultz

0.16

ÑĢÑĥÐ½

0.16

 Rating

0.15

ureka

0.15

alion

0.15

ijd

0.14

ordes

0.14

 Ratings

0.14

Activations Density 0.101%

expressions of correctness versus incorrectness in arguments

No Comments

No Known Activations