INDEX

Explanations

math and data-related words and phrases.

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

quantitative and numerical information.

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

indicators of significant or important concepts within the text, often activating for words or phrases that introduce key ideas, topics, or themes.

oai_token-act-pair · claude-3-5-sonnet-20240620 Triggered by @bot

Tokens after the letter "Q" or "M"

np_acts-logits-general · gemini-2.0-flash

word endings like ness

np_acts-logits-general · gemini-2.5-flash-lite

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 20-gemmascope-res-16k

Configuration

google/gemma-scope-2b-pt-res/layer_20/width_16k/average_l0_71

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

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

 ligiloj

-0.94

 كومونز

-0.87

AndEndTag

-0.83

sizeCache

-0.80

tvguidetime

-0.78

 متعلقه

-0.78

])).

-0.76

rungsseite

-0.75

CppMethod

-0.74

 CreateTagHelper

-0.73

POSITIVE LOGITS

ness

1.20

NESS

0.85

yet

0.80

 environment

0.77

 stuff

0.73

nesses

0.72

 parts

0.71

neſs

0.70

yet

0.68

 sounding

0.67

Activations Density 0.969%