Another #AI winter article this time from John Langford (Microsoft). Apparently I'm not isolated in my views. hunch.net/?p=9604328

John Langford on "A Real World Reinforcement Learning Research Program" -- basically laying out alternatives to the "games first, real problems later" approach to reinforcement learning research. And also a note about hiring.... 😁😁😁 hunch.net/?p=9828091

If you are interested in model-based reinforcement learning (MBRL), you want to read Iterative Value-Aware Model Learning, which is accepted at #NeurIPS2018. papers.nips.cc/paper/8121-ite…

Dimitri has a new monograph on RL and is soliciting feedback: web.mit.edu/dimitrib/www/R…

RL in the real-world: How to optimize computational pipelines on-the-fly! Eric Horvitz Besmira Nushi 💙💛 Adith Swaminathan @seanandrist alekh agarwal aditya modi Microsoft Research

In a dynamic world, static configurations are no longer enough. Researchers propose a metareasoning approach to software pipeline optimization that leverages RL to monitor pipelines and adjust module parameters on the fly for optimal performance: aka.ms/AA78h6t #AAAI20

📊 So, the results... Learning discrete distributions over a finite domain of size k to distance ε, with probability 1-δ: how hard can it be? 1/9 x.com/ccanonne_/stat…

We just scheduled Thomas Steinke (Thomas Steinke) to talk about his recent paper "Reasoning About Generalization via Conditional Mutual Information" (with Lydia Lydia Zakynthinou) on March 11! Mark your calendars, and stay tuned for further details! arxiv.org/abs/2001.09122

A new flavor of ConvNet crushes various flavors of transformers (as well as state-space models) for sequence modeling with long-range dependencies.

Come to Hall J #315 at 11a and Jinglin & Aditya Modi will tell you abt general learnability of Reward-free RL! R-f RL exhaustively explores the env & thus has heavily relied on linear structures. We now can handle non-linear FA w/ Bellman-eluder dim. More findings👇(1/2)

The 2nd Workshop on Decision Making for Modern IR and Recsys The Web Conference is calling for paper (decisionmaking4ir.github.io/WWW-2023/)! The paper submission deadline is Feb 6. #AI #ML #recsys #decisionmaking #bandits #reinforcementlearning #informationretrieval

Come, check out our poster on contextual goal-oriented RL tmr at NeurIPS!

Decision-making with LLM can be studied with RL! Can an agent solve a task with text feedback (OS terminal, compiler, a person) efficiently? How can we understand the difficulty? We propose a new notion of learning complexity to study learning with language feedback only. 🧵👇

Provably Learning from Language Feedback TLDR: RL theory can help us do better inference-time exploration with feedback. Work done with Wanqiao Xu, Ruijie Zheng, Ching-An Cheng @ICML2025, Aditya Modi, Adith Swaminathan 📰 arxiv.org/pdf/2506.10341 📍EXAIT Best Paper/Oral Sat 8:45-9:30 am

If you missed Wanqiao Xu’s presentation, here are some of our slides! (The workshop will post full slides later on their website) Paper: arxiv.org/abs/2506.10341

ICML 2017 videos have been posted icml.cc/Conferences/20…

No idea what this game is called, but whoever made it, is the devil. x.com/P_MEN876/statu…

At 4% test error, the y-axis of this plot is contained in a 95% confidence interval. Each *data point* required 450 GPUs for 7 days. x.com/OriolVinyalsML…

For my optimization followers: introducing ML to the utility of performance profiles. argmin.net/2018/03/26/per…

A contemplative article by Michael Jordan medium.com/@mijordan3/art…