INDEX

Explanations

a variety of function words in different contexts, including some medical and cultural contexts, which suggests that it recognizes the start of new sections

oai_token-act-pair · gemini-2.0-flash

descriptions in technical contexts

np_max-act-logits · gemini-2.0-flash

New Auto-Interp

Configuration

google/gemma-scope-2b-pt-transcoders/layer_22/width_16k/average_l0_15

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.22.ln2.hook_normalized

Architecture

jumprelu_transcoder

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 representations

-0.99

representation

-0.96

 Representation

-0.96

 representation

-0.96

 propOrder

-0.95

 portraying

-0.93

 Presentation

-0.91

 presentation

-0.91

 ContentValues

-0.90

 <=",

-0.90

POSITIVE LOGITS

Personensuche

0.41

rag

0.40

</i>

0.40

 extensively

0.39

 intégral

0.38

hygiène

0.36

ac

0.35

Sugar

0.34

 tarvit

0.34

 saken

0.34

Activations Density 1.649%

a variety of function words in different contexts, including some medical and cultural contexts, which suggests that it recognizes the start of new sections

descriptions in technical contexts

No Known Activations