INDEX

Explanations

negations or negative constructs in the text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 18-gemmascope-res-16k

Configuration

google/gemma-scope-2b-pt-res/layer_18/width_16k/average_l0_74

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.18.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

hatikan

-0.44

 videre

-0.42

 ändå

-0.41

 tetap

-0.40

ферен

-0.40

both

-0.39

rsiniz

-0.39

それでも

-0.38

多人

-0.38

かで

-0.38

POSITIVE LOGITS

 only

1.11

 solely

1.11

 merely

0.97

only

0.93

 onely

0.90

 Only

0.89

 seulement

0.88

 simply

0.86

Only

0.86

 ONLY

0.86

Activations Density 0.355%

negations or negative constructs in the text

No Comments

No Known Activations