format citations

joss/paper.md

@@ -55,7 +55,7 @@ We do this by introducing two reusable concepts: *a context encoder* which trans
 Our formulation unifies a wide variety of popular modeling approaches, including simple population modeling, sub-population modeling, (latent) mixture modeling, cluster modeling, time-varying models, and varying-coefficient models [@hastie1993varying], and conveniently defaults to the most appropriate type of traditional model when complex heterogeneity is not present.
 Notably, `Contextualized ML` also permits context-specific modeling even when the number of contexts vastly exceeds the number of observed samples, superceding previous frameworks by enabling even sample-specific modeling with no loss of statistical power.
 
-`Contextualized ML` is a lean, utility-oriented implementation of the broader Contextualized Machine Learning paradigm [@lengerich_contextualized_2023], focusing on the most important, novel, and popular use cases from recent works developing contextualized models [@ellington_contextualized_2023, @deuschel_contextualized_2023, @lengerich_notmad_2021, @al-shedivat_contextual_2020, @lengerich_automated_2022, @lengerich_discriminative_2020, @al-shedivat_personalized_2018].
+`Contextualized ML` is a lean, utility-oriented implementation of the broader Contextualized Machine Learning paradigm [@lengerich_contextualized_2023], focusing on the most novel and popular use cases from recent works developing contextualized models [@ellington_contextualized_2023; @deuschel_contextualized_2023; @lengerich_notmad_2021; @al-shedivat_contextual_2020; @lengerich_automated_2022; @lengerich_discriminative_2020; @al-shedivat_personalized_2018].
 We provide `Contextualized ML` as a Python package written in native PyTorch with a simple SKLearn-style interface.
 
 **Contextualized ML serves three primary purposes:**