INDEX

Explanations

directives for user interaction with content or websites

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

directs to information

np_acts-logits-general · gemini-2.5-flash-lite

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

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

ſelf

-0.74

ſelves

-0.68

 ſont

-0.65

 müſſen

-0.61

 queſta

-0.60

الحياه

-0.60

 ſich

-0.59

 increí

-0.59

AxisAlignment

-0.58

 ſur

-0.57

POSITIVE LOGITS

Kim

0.39

 irony

0.38

 references

0.37

0.36

 reference

0.35

attach

0.35

reference

0.34

BINARY

0.34

 Moulin

0.33

Activations Density 0.015%

directives for user interaction with content or websites

directs to information

No Comments

No Known Activations