HAI Industry Research

Focus:

Deep research in empathetic AI personas, long-horizon user simulation, and multi-turn workflow optimization. Investigates how conversational agents safely act with human empathy, integrate organizational context, and assist complex enterprise workflows.

Key Faculty:

James Landay, Michael Bernstein, James Zou, Monica Lam

Relevant Publications:

• HumanLM: Simulating Users with State Alignment Beats Response Imitation — Wu, Choi, Khatua, Wang, He-Yueya, Weerasooriya, Wei, D. Yang, Leskovec, Zou — ICML '26

• Bloom: Designing for LLM-Augmented Behavior Change Interactions — Jörke, Genç, Teutschbein, Sapkota, Chung, Schmiedmayer, Campero, King, Brunskill, Landay — CHI '26 (Best Paper)

• Just-In-Time Objectives: A General Approach for Specialized AI Interactions — M. Lam, Shaikh, Xu, Guo, D. Yang, Heer, Landay, Bernstein — CHI '26 (Honorable Mention)

• Controllable and Reliable Knowledge-Intensive Task-Oriented Conversational Agents with Declarative Genie Worksheets — Joshi, Liu, Chen, Weigle, Lam — ACL '25

Focus:

Cooperative multi-agent dynamics, tool integration, real-time negotiation, and front-to-back enterprise automation. Explores how autonomous systems collaborate, resolve goal conflicts, and utilize external APIs effectively.

Key Faculty:

Diyi Yang, Melissa Valentine, Tatsu Hashimoto

Relevant Publications:

• Collaborative Gym: A Framework for Enabling and Evaluating Human-Agent Collaboration — Shao, Samuel, Jiang, J. Yang, D. Yang — ICLR '2

• CollabSkill: Evaluating Human-Agent Collaboration On Real-World Tasks — Shao, Z. Wang, Ahuja, Y. Wang, Liu, D. Yang — COLM '26 (Oral)

• CooperBench: Why coding agents cannot be your teammates yet — Khatua, Zhu, Tran, Prabhudesai, Sadrieh, Lieberwirth, Yu, Fu, Ryan, Pei, D. Yang — COLM '26

• When an AI “Agentforce” enters the workforce: generative AI, employment relations, and the changing social contract — Valentine — Journal of Organization Design '26

Focus:

Advanced manipulation skills, long-horizon closed-loop control, and in-the-wild video learning. Examines how physical machines learn versatile real-world dexterity, integrate multisensory inputs, and safely interact with humans.

Key Faculty:

Shuran Song, Mykel Kochenderfer, Jeannette Bohg, Karen Liu, Jiajun Wu

Relevant Publications:

• DexUMI: Using Human Hand as the Universal Manipulation Interface for Dexterous Manipulation — M. Xu, Zhang, Hou, Z. Xu, Fan, Veloso, Song — CoRL '25 (Oral, Best Paper Finalist)

• HoMMI: Learning Whole-Body Mobile Manipulation from Human Demonstrations — Xu, Park, Zhang, Cousineau, Bhat, Barreiros, Wang, Bohg, Song — RSS '26

• Demystifying When and Why VLAs Fail in Contact-Rich Tasks and How to Fix Them — Parés Morlans, Kuhn, Liu, Longhini, Bohg — CoRL '26 (Spotlight)

• Latent Policy Barrier: Learning Robust Visuomotor Policies by Staying In-Distribution — Sun, Song — NeurIPS '25 (Spotlight)

Focus:

Next-gen architectures, decentralized LMs, TICL long-context training, and relational foundation models. Investigates scaling limits, zero-shot transfer, synthetic data scaling, and structured data representations.

Key Faculty:

Stefano Ermon, Azalia Mirhoseini, Tatsu Hashimoto, Jure Leskovec

Relevant Publications:

• Relational Transformer: Toward Zero-Shot Foundation Models for Relational Data — Ranjan, Hudovernik, Znidar, Kanatsoulis, Upendra, Mohammadi, Meyer, Palczewski, Guestrin, Leskovec — ICLR '26

• PluRel: Synthetic Data unlocks Scaling Laws for Relational Foundation Models — Kothapalli, Ranjan, Hudovernik, Dwivedi, Hoffart (SAP), Guestrin, Leskovec — ICML '26

• Synthetic Bootstrapped Pretraining — Z. Yang, Zhang, Liu, Hashimoto, Candès, Wang, Pang — ICLR '26

• Energy-Based Diffusion Language Models for Text Generation — M. Xu, Geffner, Kreis, Nie, Y. Xu, Leskovec, Ermon, Vahdat — ICLR '25

Focus:

Test-time training, error analysis, online policy gradients, and speech LMs. Develops methods for models to verify their internal reasoning steps, self-correct errors before responding, and optimize inference compute.

Key Faculty:

Percy Liang, Chris Potts, Sanmi Koyejo, Dan Jurafsky

Relevant Publications:

• s1: Simple test-time scaling — Muennighoff, Z. Yang, Shi, Li, Fei-Fei, Hajishirzi, Zettlemoyer, Liang, Candès, Hashimoto — EMNLP '25

• Outcome Rewards Do Not Guarantee Verifiable or Causally Important Reasoning — Yu, Tartaglini, Hase, Guestrin, Potts — ICML '26

• How Do Large Language Monkeys Get Their Power (Laws)? — Schaeffer, Kazdan, Hughes, Juravsky, Price, Lynch, Jones, Kirk, Mirhoseini, Koyejo — ICML '25 (Oral)

• On the Entropy Calibration of Language Models — Cao, Valiant, Liang — NeurIPS '25

Focus:

Enterprise safety standards, certified concept unlearning, factuality verification, and governance frameworks. Guarantees alignment, unlearns toxic data, and ensures reliable model behavior.

Key Faculty:

Mykel Kochenderfer, Sanmi Koyejo, Tatsu Hashimoto, Monica Lam

Relevant Publications:

• Certified Unlearning for Neural Networks — Koloskova, Allouah, Jha, Guerraoui, Koyejo — ICML '25

• An Adaptive Responsible AI Governance Framework for Decentralized Organizations — Jafari Meimandi, Reuel, Aranguiz-Dias, Rahama, Ayadi, Boullier, Verdo, Montanie (LVMH), Kochenderfer — AIES '25

• Towards Execution-Grounded Automated AI Research — Si, Z. Yang, Choi, Candès, D. Yang, Hashimoto — ICML '26

• Detecting Corpus-Level Knowledge Inconsistencies in Wikipedia with Large Language Models — Semnani, Burapacheep, Khatua, Atchariyachanvanit, Wang, Lam — EMNLP '25

Focus:

Megakernels GPU scheduling, low-resource code generation, TPU optimization, and fabric architectures. Maximizes compute throughput, reduces carbon footprint, and optimizes execution stacks.

Key Faculty:

Chris Ré, Balaji Prabhakar, Azalia Mirhoseini, Zakir Durumeric

Relevant Publications:

• ParallelKittens: Systematic and Practical Simplification of Multi-GPU AI Kernels — Sul, Arora, Spector, Ré — MLSys '26 (Oral)

• KernelBench: Can LLMs Write Efficient GPU Kernels? — Ouyang, Guo, Arora, Zhang, Hu, Ré, Mirhoseini — ICML '25

• Cartridges: Lightweight and General-Purpose Long Context Representations via Self-Study — Eyuboglu, Ehrlich, Arora, Guha, Zinsley, Liu, Rudra, Zou, Mirhoseini, Ré — ICLR '26

• Hop: A Modern Transport and Remote Access Protocol — Flammarion, Hosono, Nguyen, Bauman, Rebelsky, Wan, Adrian, Durumeric — USENIX Security '26 (Distinguished Paper runner-up)

Focus:

Stress testing, statistical arbitrage, multi-level persona generation, and risk modeling. Applies foundation models to model complex market dynamics, financial behaviors, and institutional compliance.

Key Faculty:

Renyuan Xu, Julian Nyarko, Markus Pelger

Relevant Publications:

• Deep Learning Statistical Arbitrage — Guijarro-Ordonez, Pelger, Zanotti — Management Science '26

• Tail-GAN: Learning to Simulate Tail Risk Scenarios — Cont, Cucuringu, Xu, Zhang — Management Science '26

• Attention Factors for Statistical Arbitrage — Epstein, Wang, Choi (Hanwha Life), Pelger — ICAIF '25

• Algorithmic Bias in Lending: Evidence from a Fintech Audit — Coots, Bartlett, Nyarko, Goel — Working paper '25

Focus:

AI diffusion across organizations, targeted human intervention, cognitive diversity, and labor impacts. Studies how human collaboration evolves alongside automated systems to inform organizational policy.

Key Faculty:

Melissa Valentine, Erik Brynjolfsson, Emma Brunskill, Hari Subramonyam

Relevant Publications:

• Generative AI at Work — Brynjolfsson, Li, Raymond — Quarterly Journal of Economics '25

• Future of Work with AI Agents: Auditing Automation and Augmentation Potential across the U.S. Workforce — Shao, Zope, Jiang, Pei, Nguyen, Brynjolfsson, D. Yang — Arxiv '26

• Strategic Evaluation: Incentivizing AI Capability Coverage with Private Benchmarks — Truong, S. Wang, Haber, Koyejo — NeurIPS '26

• Can LLM-Simulated Practice and Feedback Upskill Human Counselors? A Randomized Study with 90+ Novice Counselors — Louie, Shah, Orney, Pacheco, Brunskill, D. Yang — CHI '26