INDEX

Explanations

consistency and patterns in past actions or statements

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

past states or contrasts

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

ly

-0.42

recated

-0.34

is

-0.33

-0.32

rs

-0.31

gy

-0.29

 saja

-0.29

ote

-0.28

tir

-0.28

POSITIVE LOGITS

+#+

0.74

 beſch

0.68

ロウィン

0.67

 deſſen

0.66

 geſch

0.64

ðsíða

0.64

ſicht

0.63

iſche

0.63

 geweſen

0.60

 للمعارف

0.60

Activations Density 0.048%

consistency and patterns in past actions or statements

past states or contrasts

No Comments

No Known Activations