INDEX

Explanations

terms that indicate non-conditions or exceptions

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

non- prefixed 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

only

-0.37

 that

-0.36

 декабря

-0.35

Only

-0.35

that

-0.32

Dev

-0.32

Utils

-0.32

mittelt

-0.32

 июня

-0.32

But

-0.30

POSITIVE LOGITS

ſicht

0.70

tagHelperRunner

0.69

ſehen

0.69

 convertView

0.66

 CreateTagHelper

0.65

iſchen

0.63

WriteTagHelper

0.61

ſſung

0.61

 imagui

0.61

舺

0.60

Activations Density 0.033%

terms that indicate non-conditions or exceptions

non- prefixed words

No Comments

No Known Activations