Fei Fang

Associate Professor, Carnegie Mellon University

[Naveen Raman @NaveenJRaman]

A recent study found an LLM scored 95% on a healthcare benchmark. Deployed with real patients, it dropped to 34%.

In our new work, we argue the problem isn't the benchmark, but the implicit assumptions buried in evaluation.

Paper: https://t.co/mi445QtJvM

🧵 1/n

[Fei Fang @fangf07]

Excited to share this AAAI blog post on our new paper assignment algorithm used for AAAI 2026 (with Michael Cui, Chenxin Dai, and @YixuanEvenXu) and the resulting statistics. Thanks to AAAI 2026 Program Chairs Matt Taylor and Chad Jenkins, and Conference Chair @k_leyton_brown

[AAAI @RealAAAI]

Curious about the paper assignment algorithm used for AAAI 2026? The new algorithm substantially improved the robustness of large-scale paper–reviewer assignments, eliminating clear forms of strategic behavior and increasing diversity, while retaining nearly all of the assignment quality achieved by standard methods.

Read more: https://t.co/8DqCnvlIlT

[Fei Fang @fangf07]

Check out our recent work on Humanization by Iterative Paraphrasing (HIP)!

We find that commercial AI-text detectors often classify text from base LLMs as human-written, HIP leverages this observation to improve detector evasion.

[YixuanEvenXu @YixuanEvenXu]

🔗 Check out the full paper and code below:

Paper: https://t.co/i9F6kVvc2B,
Code: https://t.co/b5WgtTugAD.

Work by @YixuanEvenXu, @fjzzq2002, @AdtRaghunathan, @fangf07, @zicokolter (6/6)

[YixuanEvenXu @YixuanEvenXu]

🤖 AI text detectors are widely deployed in education and integrity workflows, but what are they actually tracking?

We report a surprising finding: text from base models is overwhelmingly judged as human by GPTZero and Pangram. 👇 (1/6)

[Fei Fang @fangf07]

Introducing our new work on interpretable AI!

[Naveen Raman @NaveenJRaman]

Training concept-based models relies on concept selection which is labor-intensive and slow.

We introduce Decision-Relevant Selection (DRS), a principled algorithm for automatic concept selection in RL.

Paper: https://t.co/TYtOJFaE4D
Website: https://t.co/NOTmQLFI8q

🧵 1/n

[Fei Fang @fangf07]

Our AI Pokémon Benchmark released!

[Seth Karten @sethkarten]

https://t.co/BxGonf2ena

[Fei Fang @fangf07]

Our recent work on AI for public mental health with @NaveenJRaman in collaboration with @hongshenus 's team, @viscidula team and @cspnj !

[CMU School of Computer Science @SCSatCMU]

SCS researchers have developed an AI-powered chatbot, PeerCoPilot, designed both with and specifically for people working in behavioral health.

👉 https://t.co/MxwTquMHfe

[CMU School of Computer Science @SCSatCMU]

SCS researchers have developed an AI-powered chatbot, PeerCoPilot, designed both with and specifically for people working in behavioral health.

👉 https://t.co/MxwTquMHfe

[YixuanEvenXu @YixuanEvenXu]

🧬 Distillation enables efficient emulation of LLMs, but verifying provenance remains a critical challenge.

Introducing Antidistillation Fingerprinting (ADFP): A principled approach that aligns signals with student learning dynamics. 👇 (1/6)

[Fei Fang @fangf07]

At 4pm, we will have our panel discussion on AI education. Panelists include our invited speakers Serene Bioth, @eunicemjun as well as Milind Tamar @MilindTambe_AI and Leo Porter

[Fei Fang @fangf07]

As a co-chair for the NeurIPS 2025 Education Program, we are excited about the One-Day Event on AI Education which will take place tomorrow Tue 2 Dec 10am - 5pm PST at Upper Level Room 9. More details here: https://t.co/PgMYxkWJQy @adityagrover_ @NaveenJRaman

[Fei Fang @fangf07]

We will have two invited talks given by Serena Booth and Eunice Jun @eunicemjun on communicating AI to non-experts.

[Fei Fang @fangf07]

As a co-chair for the NeurIPS 2025 Education Program, we are excited about the One-Day Event on AI Education which will take place tomorrow Tue 2 Dec 10am - 5pm PST at Upper Level Room 9. More details here: https://t.co/PgMYxkWJQy @adityagrover_ @NaveenJRaman

[Fei Fang @fangf07]

The first session at 10am tomorrow will be an interactive session led by Julien Besset on "Cut Through the Noise: How to Write an Effective Elevator Pitch". It will equip you with practical tools to translate your research into a short, effective, and accessible overview.

[Fei Fang @fangf07]

As a co-chair for the NeurIPS 2025 Education Program, we are excited about the One-Day Event on AI Education which will take place tomorrow Tue 2 Dec 10am - 5pm PST at Upper Level Room 9. More details here: https://t.co/PgMYxkWJQy @adityagrover_ @NaveenJRaman

[Fei Fang @fangf07]

As a co-chair for the NeurIPS 2025 Education Program, we are excited about the One-Day Event on AI Education which will take place tomorrow Tue 2 Dec 10am - 5pm PST at Upper Level Room 9. More details here: https://t.co/PgMYxkWJQy @adityagrover_ @NaveenJRaman

[Seth Karten @sethkarten]

How do we close the gap between specialist RL and generalist LLM agents?

We're benchmarking it in Pokémon. Join us at the PokeAgent Challenge competition workshop @ NeurIPS 2025.

📍 Dec 7, 8AM in San Diego
🎮 Track 1: Competitive Pokémon (game-theoretic reasoning)
🗺️ Track 2: Speedrunning (long-horizon planning)
Speakers from Google DeepMind, NYU, CMU, UT Austin, Princeton.

[Fei Fang @fangf07]

I’m so proud of you!

[Stephanie Milani @steph_milani]

📣 Honored to be selected as Honorable Mention for the @SCSatCMU Distinguished Dissertation Award!!

Thanks to my advisor @fangf07 & committee Geoff Gordon, @hongshenus, @katjahofmann, & @OriolVinyalsML (+ other mentors and collaborators) for their support 🖤

& congrats to Juncheng, Tim, and Brian 🎉

[YixuanEvenXu @YixuanEvenXu]

✨ Did you know that NOT using all generated rollouts in GRPO can boost your reasoning LLM? Meet PODS! We down-sample rollouts and train on just a fraction, delivering notable gains over vanilla GRPO. (1/7)

[Yi Wu @jxwuyi]

We release fully async RL system AReaL-boba² for LLM & SOTA code RL w. Qwen3-14B! @Alibaba_Qwen #opensource
🚀system&algorithm co-design → 2.77x faster
✅ 69.1 on LiveCodeBench
🔥 multi-turn RL ready
🔗 Project: https://t.co/YUa03ppHBp
📄 Paper: https://t.co/DMZ7YDLt6L
1/3👇

[Fei Fang @fangf07]

I’m so proud of you!

[Stephanie Milani @steph_milani]

Another life update!! 🎉

I’m joining @JHUCompSci as an Assistant Professor starting Fall 2026! Apply to work with me on reinforcement learning, foundation models, & human-centered AI. Let’s build better AI agents 🤖🙆‍♀️🦀

Before that, I’ll join @NYU_Courant as an Assistant Professor/Faculty Fellow. Excited to spend a year in NYC!

[Stephanie Milani @steph_milani]

Another life update!! 🎉

I’m joining @JHUCompSci as an Assistant Professor starting Fall 2026! Apply to work with me on reinforcement learning, foundation models, & human-centered AI. Let’s build better AI agents 🤖🙆‍♀️🦀

Before that, I’ll join @NYU_Courant as an Assistant Professor/Faculty Fellow. Excited to spend a year in NYC!

[Fei Fang @fangf07]

Excited to be at #aamas2025 !
- My keynote talk at C-MAS workshop today: 2-2:45pm, Maquette A
- Will attend panel at ALA workshop today: 4:30-5:30pm, Salon 2
- Siyu Liu (PhD advised by @___tiffanyb___ ) will present our joint paper on Friday 10:45am, Salon 3