INDEX

Explanations

instances of integration and connectedness in various contexts

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

combining various aspects

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

 Teléfono

-0.45

 chłop

-0.43

 scorso

-0.43

Quieres

-0.42

 parution

-0.42

Field

-0.42

 Defensa

-0.41

malige

-0.41

avocat

-0.39

Thursday

-0.39

POSITIVE LOGITS

 integration

0.99

 Integrate

0.98

 integrate

0.96

 Integration

0.91

 integrated

0.88

 integrates

0.87

 Integr

0.85

integration

0.85

integrate

0.84

integrated

0.83

Activations Density 0.013%

instances of integration and connectedness in various contexts

combining various aspects

No Comments

No Known Activations