INDEX

Explanations

words across medical, scientific, and technological text.

oai_token-act-pair · deepseek-r1-dist-llama-70b Triggered by @johnny

the prefix 'con' in words, often indicating the beginning of terms related to concepts, conditions, or context.

oai_token-act-pair · deepseek-v3 Triggered by @johnny

references to mathematical concepts or conjectures

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

science/tech

np_max-act-logits · gpt-4.1-nano Triggered by @johnny

con

np_max-act-logits · gemini-2.0-flash Triggered by @johnny

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 0-gemmascope-res-16k

Configuration

google/gemma-scope-2b-pt-res/layer_0/width_16k/average_l0_105

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.0.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 contextLoads

-1.07

 itſelf

-1.06

 myſelf

-0.99

 againſt

-0.94

Efq

-0.93

 becauſe

-0.92

 ſche

-0.88

 ſtate

-0.86

 Normdatei

-0.86

POSITIVE LOGITS

day

0.66

kon

0.63

per

0.60

prod

0.60

Uni

0.59

hu

0.55

kon

0.55

ual

0.54

Uni

0.52

rid