INDEX

Explanations

references to covert operations, particularly those involving deception and planned attacks

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

references to violent or covert actions like assassination, plotting, and military/intelligence operations.

oai_token-act-pair · claude-3-5-haiku-20241022 Triggered by @neilrathi

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B @ 20-gemmascope-res-16k

Configuration

google/gemma-scope-9b-pt-res/layer_20/width_16k/average_l0_68

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.20.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

 المعيارى

-0.63

 initComponents

-0.56

setLayoutManager

-0.52

 desmotivaciones

-0.48

 zelve

-0.45

 digitais

-0.44

 Económica

-0.44

 ziekte

-0.44

 dezelve

-0.44

ciller

-0.43

POSITIVE LOGITS

 stealth

0.51

UserScript

0.50

 infiltration

0.48

wapV

0.48

 sabotage

0.47

 sneaking

0.46

 clandestine

0.45

 sneak

0.45

 infiltrate

0.45

 cland

0.44

Activations Density 0.224%

references to covert operations, particularly those involving deception and planned attacks

references to violent or covert actions like assassination, plotting, and military/intelligence operations.

No Comments

No Known Activations