INDEX

Explanations

numerical values and options in a list format

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

numbers and words

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

]")]

-0.56

RenderAtEndOf

-0.56

jspb

-0.55

 administracy

-0.54

期刊论文

-0.52

InjectAttribute

-0.52

migrationBuilder

-0.52

 Paglinawan

-0.51

majánló

-0.51

 snippetHide

-0.51

POSITIVE LOGITS

 MIDDLE

0.45

 Middle

0.44

Middle

0.41

 word

0.39

 middle

0.37

Custom

0.37

lighet

0.36

↵↵↵

0.35

word

0.35

 custom

0.34

Activations Density 0.010%

numerical values and options in a list format

numbers and words

No Comments

No Known Activations