INDEX

Explanations

items that denote pauses or breaks in thought or conversation, particularly punctuation marks such as commas

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

comma followed by "what", "there", "come", "stick", "sis", "пожалуйста", "really", "and", "coffee", "No", "knowing", "we"

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

-0.41

 following

-0.38

 Following

-0.36

 themselves

-0.35

 suivants

-0.35

Following

-0.34

 vêtement

-0.33

׃

-0.33

 desplazamiento

-0.33

 himself

-0.32

POSITIVE LOGITS

⟬

0.68

ftagPool

0.68

 kasarigan

0.62

tvguidetime

0.60

.$,

0.59

saraba

0.58

readystatechange

0.57

,<

0.57

cticut

0.57

ktop

0.56

Activations Density 0.073%

items that denote pauses or breaks in thought or conversation, particularly punctuation marks such as commas

comma followed by "what", "there", "come", "stick", "sis", "пожалуйста", "really", "and", "coffee", "No", "knowing", "we"

No Comments

No Known Activations