INDEX

Explanations

programmatic structures and components, particularly related to classes and APIs in code

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

programming class definitions

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

+#+

-0.90

 estekak

-0.77

 nahilalakip

-0.73

 queſta

-0.73

CppMethod

-0.70

 defaultstate

-0.69

Tikang

-0.69

 متعلقه

-0.68

[@BOS@]

-0.68

<unused41>

-0.68

POSITIVE LOGITS

Com

0.36

Per

0.35

Per

0.31

cont

0.28

PER

0.28

Consider

0.28

is

0.27

0.26

Con

0.26

↵↵

0.26

Activations Density 0.006%

programmatic structures and components, particularly related to classes and APIs in code

programming class definitions

No Comments

No Known Activations