INDEX

Explanations

references to academic titles and affiliations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 29-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L29R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.29.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

anela

-0.17

_dispatch

-0.15

isted

-0.15

PCA

-0.15

 Trident

-0.14

 Watson

-0.14

 Straw

-0.14

 Ranger

-0.14

acea

-0.14

ld

-0.14

POSITIVE LOGITS

 Sunny

0.28

 Sick

0.27

Cum

0.19

 sick

0.18

 Sinai

0.17

ICES

0.17

CI

0.17

Cum

0.16

.ci

0.16

 Schul

0.16

Activations Density 0.007%

references to academic titles and affiliations

No Comments

No Known Activations