📢 Thrilled to announce our latest paper! "Merge Conflicts!" Exploring the Impacts of External Distractors to Parametric Knowledge Graphs: qiancheng0.github.io/files/Impact_o… 🤖 Experience the CLASH of external knowledge and LLM's parametric knowledge through its inner mechanism! (1/n)

🚨 Reward Models in RLHF are trained to reflect human preference, but can they consistently do so in practice? We study the phenomenon of reward inconsistency and investigate its impact on the RLHF process. arxiv.org/pdf/2309.16155… Work led by @Shadowkiller331! 🧵1/10

❗With dense retrieval, the unit in which you segment a retrieval corpus (passage, sentence, etc) may impact performance by more than you thought! We introduce a novel retrieval unit, proposition, for dense retrieval. chentong0.github.io/factoid-wiki/ [1/7]

Excited to share the code release of RAPTOR - you can get started with just a few lines of code, so check it out!

🌟 Want your own private "autopilot" system? Check out our newly released Cognitive Kernel system! 🤖 This powerful tool could handle daily tasks involving: 🌐 Real-time information 📁 Private files 🕰️ Long-term memory Plus, it's fully dockerized for easy local use! 🚀

In 2020, we introduced an AI method capable of generating superhuman chip layouts. Today, we describe its impact on the field and give it a name: AlphaChip!

🎯Perspective-aware IR IR is crucial for RAG. Existing works solve general IR tasks pretty well, but what about real-world tasks requiring retrieval from multiple perspectives?🤔 We’ll present our "PIR" work at the #COLM2024 poster session tomorrow morning. Come check it out!!

Dense X Retrieval: What Retrieval Granularity Should We Use? by Tong Chen, Hongwei Wang, Sihao Chen, Wenhao Yu, Kaixin Ma, Xinran Zhao, Dong Yu, and Hongming Zhang Session: Information Retrieval and Text Mining 1, Session 02, 11:00-12:30 aclanthology.org/2024.emnlp-mai…

MixGR: Enhancing Retriever Generalization for Scientific Domain through Complementary Granularity by Fengyu Cai, Xinran Zhao, Tong Chen, Sihao Chen, Hongming Zhang, Iryna Gurevych, & Heinz Koeppl Dialogue & Interactive Systems 3, Session 12, 14:00-15:30 aclanthology.org/2024.emnlp-mai…

Overall, our work serves to challenge n-gram data membership definition in LLMs, and call for better definitions that capture human intuition as it underpins many important topics today. Paper: arxiv.org/abs/2503.17514 A collab of Stanford AI Lab, @StanfordNLP, and

John Zimmerman presenting at #CHI2025 Catch him and Ask about our AI Literacy paper — best paper honorable mention Motahhare Eslami We thank Ken Holstein Ken Koedinger Amy Ogan Howard(Ziyu) Han Yanlin Du for their feedback on this journey. programs.sigchi.org/chi/2025/progr…

CLS Through reading them, I feel I can roughly tell which ones are AI-generated. Is it possible to use AI-generated content as secret sanity-check questions used in surveys? Organizers know which are AI-generated, and reviewers need to flag them to reflect the review quality

Spurious Rewards was not all‼️We now present spurious PROMPTS🤔 check out our latest findings and discussion on evaluation: tinyurl.com/spurious-prompt. Who knew Lorem ipsum can bring 19.4% gains compared to default prompt👀 Also, arXiv is out🤩 arxiv.org/abs/2506.10947📄

Your bimanual manipulators might need a Robot Neck 🤖🦒 Introducing Vision in Action: Learning Active Perception from Human Demonstrations ViA learns task-specific, active perceptual strategies—such as searching, tracking, and focusing—directly from human demos, enabling robust

Revela: Dense Retriever Learning via Language Modeling Introduces a unified framework for self-supervised retriever learning via language modeling with in-batch attention mechanism. 📝arxiv.org/abs/2506.16552 👨🏽‍💻github.com/TRUMANCFY/Reve…

MoR: Better Handling Diverse Queries with a Mixture of Sparse, Dense, and Human Retrievers Jushaan Kalra et al. present a zero-shot framework that dynamically combines heterogeneous retrievers for each query. 📝arxiv.org/abs/2506.15862 👨🏽‍💻github.com/Josh1108/Mixtu…

Are AI scientists already better than human researchers? We recruited 43 PhD students to spend 3 months executing research ideas proposed by an LLM agent vs human experts. Main finding: LLM ideas result in worse projects than human ideas.