INDEX

Explanations

frequency-related adverbs indicating how often actions occur

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B @ 2-gemmascope-res-16k

Configuration

google/gemma-scope-9b-pt-res/layer_2/width_16k/average_l0_67

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.2.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 XNUMX

-0.59

 EconPapers

-0.57

Hochspringen

-0.56

 nahilalakip

-0.54

 quase

-0.54

DotNetBar

-0.53



-0.53

IContainer

-0.53

verifyException

-0.52

 HttpNotFound

-0.50

POSITIVE LOGITS

 specifically

0.46

 beforehand

0.45

uestamente

0.44

 giản

0.43

 differently

0.43

 regularly

0.43

 correctly

0.42

 урна

0.41

 sooner

0.40

 itself

0.40

Activations Density 0.332%

frequency-related adverbs indicating how often actions occur

No Comments

No Known Activations