Kaixiang

RL researcher. Opinions are my own.

[Kaixiang @linkaixi]

Attending #ICLR2025 from 4/23 to 4/28 & will present PrefEval (https://t.co/9WfE0IwAvz) discussing the performance of SoTA LLMs on personalization.

Catch our presentations on April 26th:
🔥Oral: 10:42-10:54am @ Hall 1 Apex
📊 Poster #558: 3:00-5:30pm @ Hall 3 + Hall 2B

Beyond personalization, If you’d like to chat about RL, reasoning, & multimodal agents. DM for a coffee chat, let's connect! #ICLR2025 #RL #LLM #AmazonScience

[Siyan Zhao @siyan_zhao]

Excited to release PrefEval (ICLR '25 Oral), a benchmark for evaluating LLMs’ ability to infer, memorize, and adhere to user preferences in long-context conversations!

⚠️We find that cutting-edge LLMs struggle to follow user preferences—even in short contexts. This isn't just about long-context ability—but also a lack of proactiveness in following preferences.

💭 LLMs struggle with implicit preferences revealed through conversation, requiring more reasoning.

🚀 SFT on our benchmark greatly improves performance w/ enhanced attention to preference regions!

🧩 What's in PrefEval?
- 3,000 user preference & query pairs
- 3 Implicit & explicit preference forms
- 20 conversational topics
- Evaluated up to 100K context w/ various baselines
- Flexible task forms: MCQ & generation
- Extensive Error Type Analysis

🔗 Learn more:
Website: https://t.co/nwSEbefqFf
Code & Data: https://t.co/5h1SphruHN
Paper: https://t.co/J3bejwsIpq

Huge thanks to my mentors during my internship at Amazon @linkaixi @Mingyi552237, Yang Liu, @hdevamanyu!

[Yifei Zhou @YifeiZhou02]

🚨🚨🚨 What to do when pre-training ends? Excited to share our latest work Proposer-Agent-Evaluator (PAE), where we trained an open-source SOTA generalist VLM web agent entirely with self-generated data and autonomous RL.

Infrastructure and model fully open-sourced.

(1/9)

[Kaixiang @linkaixi]

@_zhiweijia @govind_thattai @gauravsukhatme @qiaozikl @AmazonScience

[Zhiwei Jia @_zhiweijia]

I am presenting a paper from my last year's Amazon internship at IROS 2022, in which we handle long-horizon planning for home-assistant virtual robots with better semantic representation and exploration strategy.

https://t.co/FmPVvikmDa

#IROS2022

[Huaxiu Yao @HuaxiuYaoML]

I'm on the academic job market this year!
I develop machine learning methods that are robust and adaptable to distribution shifts and open/non-stationary environments, w/ interdisciplinary applications in healthcare+drug discovery, transportation, and education. RT appreciated.

[Amazon Science @AmazonScience]

Amazon and @UCLA announced the establishment of the Science Hub for Humanity and Artificial Intelligence. The collaboration will support research, education & outreach efforts in areas of mutual interest around #AI and its applications to benefit humanity. https://t.co/iGnng6Du3I

[Hua Wei@ICML @realhuawei]

A late personal update: I'm now a tenure-track assistant professor at the New Jersey Institute of Technology (NJIT). Fresh new start! And I'm looking for prospective students interested in RL/DM/Transportation for Spring/Fall 2022. https://t.co/UySNQ67gku

[Tyler Derr @TylersNetwork]

🚨Postdoc opportunity🚨 in my group @nds_vu on topics in social network analysis!
Please get in touch with me via email asap!

We especially encourage applicants who are members of underrepresented groups (including those who identify as neurodivergent).

📢 Please RT!

[Tyler Derr @TylersNetwork]

Please see the link below if seeking a PhD position starting Fall 2021 or Spring 2022 in “...data mining and machine learning research, especially in graph neural networks and social network analysis.”

You can follow Yao @ma_yao_

https://t.co/8yZI1qM799

[Amazon Science @AmazonScience]

“Despite the challenges of the pandemic, the Alexa team has shown incredible adaptability and grit, delivering scientific results that are already making a difference for our customers and will have long-lasting effects.” Rohit Prasad, VP & Head Scientist. https://t.co/Y3wTXKq5Af

[Xiaorui Liu @liu_xiaorui]

Excited to share that our paper "Linear Convergent Decentralized Optimization with Compression" has been accepted in ICLR 2021 https://t.co/pxX0ibAYsC Thank the reviewers/ACs and our collaborators @tangjiliang, @MingYan08, Yao Li, Rongrong Wang

[Xikun Zhang 张熙堃 @xikun_zhang_]

Very happy to have my previous collaborator @iftenney at @GoogleAI to give a talk at the Stanford NLP seminar this Thursday! Gonna be exciting!

[Microsoft Research @MSFTResearch]

Reinforcement Learning Day 2021 will feature a lively debate between Prof Yoshua Bengio from @Mila_Quebec and Dr. @JohnCLangford. Reserve your seat now to watch their discussion on “The State of RL and The Theory-Practice Divide” on January 14 at 11 AM: https://t.co/mZJIG6rHYw

[Wei Jin @weisshelter]

📣[DeepRobust-0.1.1 Comes Out!] Our PyTorch library for adversarial robustness on graph and image data, DeepRobust, can now be installed from pip! Just try `pip install deeprobust`. For more details, please visit https://t.co/ulwB3YAVaU! #MachineLearning #python #DeepLearning

[Kaixiang @linkaixi]

What a comprehensive collection of simulation platforms for embodied AI and RL! Add two more: https://t.co/MOUHtwnhm7 and https://t.co/w4Ud9C3tcq

[Ankur Handa @ankurhandos]

Some simulation platforms that caught my eye this year 🧵

[Kaixiang @linkaixi]

Embodied & Teachable AI team at Amazon Alexa AI is looking for multiple research interns starting summer 2021.

Please feel free to contact me if you are interested in Human-robot interaction, VQA, VLN, etc. #internship #machinelearning #AI #Amazon

[Kaixiang @linkaixi]

#SIAMSDM18 Markov Chain Monitoring by Harshal Chaudhari et al. Design efficient algorithm to reduce the variance of prediction of positions of items with limited monitoring resources

[Kaixiang @linkaixi]

#SIAMSDM18 Co-Regularized Monotone Retargeting for Semi-Supervised Letor by Shalmali Joshi et al. Utilizing convex structures of isotonic vectors for a semi-supervised learning to rank problem under multi-view setting.

[Kaixiang @linkaixi]

#SIAMSDM18 Learning Graph Representation Via Frequent Subgraphs by Dang Nguyen. Unsupervised learning for graph embedding.