INDEX

Explanations

terms indicating high-quality or superior classes and grades

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

 Navidad

-0.45

FACEBOOK

-0.42

 doença

-0.38

 nevera

-0.37

 Asamblea

-0.37

Jahres

-0.37

Reducers

-0.36

 mensajes

-0.36

 regalos

-0.36

Obec

-0.36

POSITIVE LOGITS

 quality

0.86

quality

0.86

Excellent

0.80

excellent

0.77

 excellent

0.77

Quality

0.76

 Excellent

0.75

notch

0.75

 excellence

0.75

 Quality

0.74

Activations Density 0.012%

terms indicating high-quality or superior classes and grades

No Comments

No Known Activations