Meng Chen

phd @Berkeley_EECS • interests={human-centered ai, collaborative agents} • en/zh

[Meng Chen @mengchen_24]

Happening soon in Continental Ballroom 4!

[Meng Chen @mengchen_24]

I'm going to #COLM2026!
I’ll present DigitalCoach as an 💡oral spotlight (Top 3%) at the Workshop on Agent Behavior (https://t.co/Y0g4kXxuLs) on Oct 9 (1- 2 p.m.)!
I’d love to connect with folks in industry and academia working on human-agent collaboration and user simulation!

[Philippe Laban @ COLM @PhilippeLaban]

I'm at COLM this week, presenting:
- LLMs Corrupt Your Documents When You Delegate. At today (Wednesday's) afternoon poster 4.30-6.30. Come check out our very (!) colorful poster.

- Giving a fun talk on mostly ongoing work at the Workshop on Lifelong Agents early Friday morning.

Old and new friends, let's chat!

[Meng Chen @mengchen_24]

I'm going to #COLM2026!
I’ll present DigitalCoach as an 💡oral spotlight (Top 3%) at the Workshop on Agent Behavior (https://t.co/Y0g4kXxuLs) on Oct 9 (1- 2 p.m.)!
I’d love to connect with folks in industry and academia working on human-agent collaboration and user simulation!

[Meng Chen @mengchen_24]

Excited to share that our paper 𝐃𝐢𝐠𝐢𝐭𝐚𝐥𝐂𝐨𝐚𝐜𝐡 was accepted to the #EMNLP2026 𝐌𝐚𝐢𝐧! 🎉

Agents are getting better at automating software tasks. But can they teach us how to use software ourselves?

We studied real expert–novice coaching interactions: 🧵

[Meng Chen @mengchen_24]

@jiegaosophia I would also be at colm!!

[Meng Chen @mengchen_24]

@jiegaosophia Or we can build a machine to help us better at these ;)

[Meng Chen @mengchen_24]

Can’t agree more.
My recent obsession: If AI makes expert execution abundant, the bottleneck moves upstream.
Goal formulation, taste, and judgment may become more important than ever. But then: Can we build a machine to do that too? 🤔

[lencx @lencx_]

AI 是放大器，不是许愿池。模型可以帮你跨过操作门槛，却不能凭空补上领域知识，也不会自动把模糊的审美变成清晰的标准。 https://t.co/xeCzzDyNaH

[Meng Chen @mengchen_24]

@realJessyLin wow -- which book is this?

[Meng Chen @mengchen_24]

@FINeuraLab indeed. that will show a clearer picture of whether the coaching agent is reciting the steps or actually giving adaptive instructions

[Meng Chen @mengchen_24]

Excited to share that our paper 𝐃𝐢𝐠𝐢𝐭𝐚𝐥𝐂𝐨𝐚𝐜𝐡 was accepted to the #EMNLP2026 𝐌𝐚𝐢𝐧! 🎉

Agents are getting better at automating software tasks. But can they teach us how to use software ourselves?

We studied real expert–novice coaching interactions: 🧵

[Meng Chen @mengchen_24]

@lihanc02 Dataset opensource now

[Meng Chen @mengchen_24]

📄𝐏𝐚𝐩𝐞𝐫: https://t.co/1MkTOvk6Xi
💻𝐏𝐫𝐨𝐣𝐞𝐜𝐭 𝐏𝐚𝐠𝐞: https://t.co/irPoc4h32O
🤗𝐃𝐚𝐭𝐚: https://t.co/IKpnhe1YAz
Huge thanks to my wonderful collaborators @__anyaj, @tsunghan_wu, Tobias Maringgele, @_dmchan, @alsuhr, and @amypavel
(n/n)

[Meng Chen @mengchen_24]

This points to two exciting directions:

𝐀𝐠𝐞𝐧𝐜𝐲-𝐩𝐫𝐞𝐬𝐞𝐫𝐯𝐢𝐧𝐠 𝐚𝐠𝐞𝐧𝐭𝐬 that help people build skills and act independently.

𝐒𝐜𝐚𝐥𝐚𝐛𝐥𝐞 𝐢𝐧𝐭𝐞𝐫𝐚𝐜𝐭𝐢𝐯𝐞 𝐞𝐯𝐚𝐥𝐮𝐚𝐭𝐢𝐨𝐧 beyond single turns, measuring how interactions change human outcomes. (6/n)

[Meng Chen @mengchen_24]

📄 𝐏𝐚𝐩𝐞𝐫: https://t.co/1MkTOvk6Xi
💻 𝐏𝐫𝐨𝐣𝐞𝐜𝐭 𝐏𝐚𝐠𝐞: https://t.co/qocmvPZj3o
🤗 𝐃𝐚𝐭𝐚: https://t.co/IKpnhe1YAz

Huge thanks to my wonderful collaborators @__anyaj, @tsunghan_wu, Tobias Maringgele, @_dmchan, @alsuhr, and @amypavel 🙌 (n/n)

[Meng Chen @mengchen_24]

This points to two exciting directions:

𝐀𝐠𝐞𝐧𝐜𝐲-𝐩𝐫𝐞𝐬𝐞𝐫𝐯𝐢𝐧𝐠 𝐚𝐠𝐞𝐧𝐭𝐬 that help people build skills and act independently.

𝐒𝐜𝐚𝐥𝐚𝐛𝐥𝐞 𝐢𝐧𝐭𝐞𝐫𝐚𝐜𝐭𝐢𝐯𝐞 𝐞𝐯𝐚𝐥𝐮𝐚𝐭𝐢𝐨𝐧 beyond single turns, measuring how interactions change human outcomes. (6/n)