INDEX

Explanations

frequent occurrences of specific characters and symbols used in programming or code structures

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

code identifiers like NGINX, HTTP, or types like `DockStyle`

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

 kyse

-0.45

 Wilayah

-0.45

 circuitos

-0.44

 berdua

-0.44

 ακόμα

-0.43

StatusOK

-0.43

مراجع

-0.43

 Gutes

-0.43

Референце

-0.43

 cuantos

-0.43

POSITIVE LOGITS

RTEE

0.56

DockStyle

0.56

providedIn

0.53

UserScript

0.51

setVerticalGroup

0.49

المكان

0.45

webElementXpaths

0.44

HtmlAttribute

0.44

 EconPapers

0.43

migrationBuilder

0.42

Activations Density 0.024%

frequent occurrences of specific characters and symbols used in programming or code structures

code identifiers like NGINX, HTTP, or types like `DockStyle`

No Comments

No Known Activations