INDEX

Explanations

phrases indicating actions and their outcomes in various contexts

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

past numbers and percentages

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

sproz

-0.40

 Tuhan

-0.40

 countries

-0.40

 gouver

-0.40

uramente

-0.39

 certaine

-0.38

</thead>

-0.38

 habitudes

-0.38

 oignon

-0.37

 certain

-0.37

POSITIVE LOGITS

 nakalista

0.60

setVerticalGroup

0.59

 snippetHide

0.53

pgterms

0.51

tagHelperRunner

0.50

 HttpHeaders

0.50

encodeWith

0.49

BeginContext

0.48

HtmlAttribute

0.48

----</

0.48

Activations Density 0.113%

phrases indicating actions and their outcomes in various contexts

past numbers and percentages

No Comments

No Known Activations