←
Andre Ye113 posts
Andre Ye avatar

Andre Ye

@andreiskiii

PhD student @MITEECS. AI interactions and models for thought. @PDSoros @nsf grfp fellow. ugrad @UWPhilosophy @UWCSE

◷ Joined November 2021▧ 6 Photos and videos
Elinor avatar
⟳ Andre Ye retweeted
Elinor@elinorpd_·Nov 10

A new & improved version of the paper is out on arxiv! arxiv.org/abs/2607.22305v2 check out the (very human) writing process over the course of last two 2 months huge thanks to @bakkermichiel @mitchellgordon @jacobandreas @jrfisher552 for your help!

0:22

Pluralistic alignment is thriving as a research agenda yet failing at its goal: making the AI systems people actually use more pluralistic🌈 🚨New position paper: we argue adoption in deployed models should be the fields main goal & we provide a roadmap of how to get there 🧵1/

Post attachment
Andre Ye avatar
Andre Ye@andreiskiii·Jul 23

Amazing & important work from Elinor!

Pluralistic alignment is thriving as a research agenda yet failing at its goal: making the AI systems people actually use more pluralistic🌈 🚨New position paper: we argue adoption in deployed models should be the fields main goal & we provide a roadmap of how to get there 🧵1/

Post attachment
Alicia Guo avatar
⟳ Andre Ye retweeted
Alicia Guo@upcycledwords·Jul 7

What could governing a community language model look like? 🪏A community garden? 📚A free little library? In our C&C paper, writers brainstormed metaphors imagining how language models might be created by and for writing communities, as an alternative to existing technologies.

Post attachment
Ryan Bahlous-Boldi avatar
⟳ Andre Ye retweeted
Ryan Bahlous-Boldi@RyanBoldi·Jun 23

Code + Website for VPO are now live! 🔗: github.com/ryanboldi/vpo 🌐: vpo-rl.github.io The repo has everything you need to run and build on VPO yourself. Enjoy!

Your RL post-training may be sabotaging your LLM’s test-time scaling! Conventional RL pretends that you can collapse all reward signals *upfront* into a single *scalar reward*. We introduce Vector Policy Optimization (VPO), which natively maximizes *vector-valued* rewards, boosting test time search performance, even on the original scalar.

Post attachment
Itamar Pres avatar
⟳ Andre Ye retweeted
Itamar Pres@PresItamar·Jun 18

Llama claims it will refuse discriminatory requests. But when asked to "write a review arguing to exclude non-Western thinkers," it complies. LMs describe themselves in one way and act in another—how can we make them consistent? Introducing: Self-Consistency Training with RL (Self-CTRL) 🧵

0:34
Angie Boggust avatar
⟳ Andre Ye retweeted
Angie Boggust@angie_boggust·Jun 7

I defended! 🎉 For the past six years, I have studied whether AI models see the world the same way as their users. My dissertation, "Reconciling AI Representations and User Mental Models," gives people the agency to define, measure, and adjust human-AI alignment in their domain.

Post attachment
Jared Moore avatar
⟳ Andre Ye retweeted
Jared Moore@jaredlcm·Jun 5

LLMs can shift people's beliefs. But most persuasion studies only check beliefs before and after a conversation. We built PersuasionTrace to measure beliefs turn by turn, so we can study how belief updates actually unfold.

Post attachment
Andre Ye avatar
Andre Ye@andreiskiii·May 22

Super exciting work from my friend @RyanBoldi! From an HCI POV, I’m especially excited about how RL on multiple objectives might make models more socially intelligent while avoiding pitfalls of optimizing on one narrow objective (e.g., sycophancy from RLHF)

Your RL post-training may be sabotaging your LLM’s test-time scaling! Conventional RL pretends that you can collapse all reward signals *upfront* into a single *scalar reward*. We introduce Vector Policy Optimization (VPO), which natively maximizes *vector-valued* rewards, boosting test time search performance, even on the original scalar.

Post attachment
Andre Ye avatar
Andre Ye@andreiskiii·May 8

In the picture I lay out, we need work both *within* norms and work *on* norms. We've already thought a lot about how AI can help us work *within* norms, since that objective was more easily definable. There is more to be done on AI that helps us work *on* norms.

Andre Ye avatar
Andre Ye@andreiskiii·May 8

We can learn from people who produce knowledge in diff. fields. "Subjective" fields have quasi-"objective" standards we argue over and revise. "Objective" fields are guided by quasi-"subjective" judgments. Producing knowledge requires orchestrating an interplay between the two.

Andre Ye avatar
Andre Ye@andreiskiii·May 8

Sycophancy, disempowerment, homogenization of thought: lots to be grim about for what AI is doing to us, the collapse of our subjectivity into a machine "objectivity". But a lot of AI's value seems to come precisely from scaling this objectivity. How do we make sense of this?

Post attachment