To Understand Language is to Understand Generalization | Eric Jang

In my essay “Just ask for Generalization”, I argued that some optimization capabilities, such as reinforcement learning from sub-optimal trajectories, might be better implemented by generalization than by construction. We have to generalize to unseen situations at deployment time anyway, so why not focus on generalization capability as the first class citizen, and then “just ask for optimality” as an unseen case? A corollary to this design philosophy is that we should discard inductive biases that introduce optimization bottlenecks for “the data sponge”: if an inductive bias turns out to be merely “data in disguise”, it may not only cease to provide a benefit in the high data regime, but actually hinder the model on examples where the inductive bias no longer applies.

In my essay "Just ask for Generalization" , I argued that some optimization capabilities, such as reinforcement learning from sub-optimal trajectories, might be better implemented by generalization than by construction . We have to generalize to unseen situations at deployment time anyway, so why not focus on generalization capability as the first class citizen, and then "just ask for optimality" as an unseen case? A corollary to this design philosophy is that we should discard inductive biases that introduce optimization bottlenecks for "the data sponge": if an inductive bias turns out to be

saved by

• Winnie Xu
• Alicia Guo
• Joshua Baek
• Jim Wu
• Stephen Fay
• SUJASH AGRAWAL

related reading

• Just Ask for Generalization | Eric Jangevjang.com
• How Can We Make Robotics More like Generative Modeling? | Eric Jangevjang.com
• Language model harnesses are compositional generalizersalexzhang13.github.io
• Understanding deep learning requires rethinking generalizationarxiv.org
• Why Does Reinforcement Learning Generalize? A Feature-Level Mechanistic Study of Post-Training in Large Language Modelsarxiv.org
• Do Machine Learning Models Memorize or Generalize?pair.withgoogle.com
• Do large language models understand us? | by Blaise Aguera y Arcas | Mediummedium.com
• A History of Large Language Modelsgregorygundersen.com
• Taking LLMs Seriously (As Language Models) — LessWronglesswrong.com
• What language do language models speak?tcz.hu
• [2503.02113] Deep Learning is Not So Mysterious or Differentarxiv.org
• Generalization Dynamics of LM Pre-training — Jiaxin Wenjiaxin-wen.github.io