INDEX

Explanations

phrases advocating for suggestions or recommendations

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

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 myö

-0.47

 sitten

-0.47

 podjet

-0.41

 instituição

-0.39

 cualquiera

-0.38

 paciente

-0.38

 saira

-0.38

 aislado

-0.37

 enää

-0.37

 enjeux

-0.37

POSITIVE LOGITS

 recommended

1.09

Recommended

1.03

 Recommended

1.02

recommended

0.99

RECOMMENDED

0.90

 recommend

0.85

 RECOMMENDED

0.83

Suggested

0.82

 Suggested

0.79

 recommends

0.79

Activations Density 0.006%

phrases advocating for suggestions or recommendations

No Comments

No Known Activations