Paul Liang

This account is verified. Learn more

@pliang279

Assistant Professor MIT @medialab @MITEECS @nlp_mit || Foundations of self-evolving multisensory AI to enhance the human experience.

pliang279.github.io Joined July 2012

464 Following 12K Followers 3.8K Posts

Latest posts 5

Pinned post

Paul Liang @pliang279 Oct 22

The next frontier is AI that improves itself, and we believe everyone should know how to build it.

This semester, Ao @ao_qu18465 and I are teaching 'How to AI an AI for Almost Anything' @MIT @medialab -- a new course on self-evolving AI systems that recursively improve from experience, feedback, memory, and exploration over long interaction horizons.

The key idea: intelligence shouldn't stop improving when training ends.

We cover four levels of evolution:
(1) context evolution through prompt optimization and context management,

(2) harness evolution through memory, tools & skills, workflow optimization, and multi-agent systems,

(3) model evolution through RL, distillation, test-time training, self-play, and continual learning,

(4) task evolution through generating increasingly complex synthetic tasks, simulations, evaluations, and environments.

Slides, readings, and videos available: mit-mi.github.io/how2ai-course/…

We thank @Kimi_Moonshot and @PrimeIntellect for generously sponsoring API credits!

Paul Liang @pliang279 Oct 22

I am particularly excited about SteerAct — by just steering **one** neuron, with 24 task demos, we take a frozen VLA/WAM and adapt it to a variety of downstream robotic tasks, with large gains in performance.

see paper and code here: steeract.github.io

and full thread by @alex_kai2020

kaichen_zhou @alex_kai2020 Oct 20

🚀SteerAct: Stanford × MIT x Harvard
🧠24 test-time rollouts/task. Only steering one feature. 🔒Frozen VLA/WAM.
🤖Real-robot mean success: 31.1% → 52.2% (+21.1 pp; ~68% relative).
🎬See it in action: arxiv.org/pdf/2610.25632
@MIT @medialab @Harvard @Stanford
#Robot #AI

View quoted post on X

Paul Liang @pliang279 Oct 6

introducing Guidance-TTT - evolving a small 8B model is sufficient to guide discovery, outperforming significantly larger models!

test time training of small idea generating models (ie, with research taste) will be increasingly important.

see code and models on our open-source infrastructure reef: reefinfra.ai/docs/user-guid…

paper: arxiv.org/abs/2610.06269

www.reefinfra.ai Guidance-TTT: train the guide, keep the executor frozen Guidance-TTT trains a small model to propose improvements to a candidate program. A frozen execution model turns each proposal into code. A task judge evaluates that code, and Reef uses the score to...

Ao Qu Verified account @ao_qu18465 Oct 6

(1/8) 💡 To make better discoveries, it can be enough to train only the idea generator.

Guidance-TTT trains a small 8B model at test time to propose better ideas, while a frozen model turns them into code. The resulting programs are then evaluated to provide the reward for updating the small model. 🔄

Using only open-weight models and no web access, we outperform the best reported results on 4 discovery tasks. 🧵

Paper and code are out:

📄 Paper: arxiv.org/abs/2610.06269

💻 Code: github.com/Human-Agent-So…

View quoted post on X

Paul Liang @pliang279 Sep 26

Update your agent just by talking to it!

Reef introduces the /reefine command: just describe what you want your agent to do. Reef builds the change, checks it, and ships a new version of the agent with an updated harness.

Examples:
💬 "> /reefine add a /chat mode for faster responses"
🔊 "> /reefine tell me out loud when you're done"
⚡ "> /reefine add jev as a tool": 120 papers screened in 9 s

Quick and personalized harness evolution will be key for continual learning and self-improving agents.

Reef also supports many other functions, including test-time training and model evolution. Try it here: github.com/Human-Agent-So…

github.com GitHub - Human-Agent-Society/reef: Infrastructure for continually self‑improving agents Infrastructure for continually self‑improving agents - Human-Agent-Society/reef

Ao Qu Verified account @ao_qu18465 Sep 25

Your agent can now grow new abilities, just by asking.

Reef now supports personalized harness evolution with /reefine: describe what you want your agent to do. Reef builds the change, checks it, and ships a new version.

Examples:
💬 "> /reefine add a /chat mode for faster responses"
🔊 "> /reefine tell me out loud when you're done"
⚡ "> /reefine add jev as a tool": 120 papers screened in 9 s

Works with any agent harness through a Reef adapter: Codex, Hermes, OpenCode, Pi and Terminus 2 today.

Try it here: github.com/Human-Agent-So…

#agent #harness #rsi #llm

View quoted post on X

Paul Liang @pliang279 Sep 7

Sharing lecture videos for the **How to AI (Almost) Anything/Multimodal AI** course I taught at MIT in spring 2026.

This course became quite a hit the last time I shared it in spring 2025. Spring 2026's updated version contains updated topics on multimodal agents, reasoning, self-evolving AI, and new modalities like touch and smell. Also includes slides (not videos) of guest lectures on multimodal AI for health, design, manufacturing, cities, & transportation.

Youtube playlist: youtube.com/playlist?list=…

Course website and materials: mit-mi.github.io/mmai-course/sp…

Today's AI can be applied to almost anything - from language to vision, audio, sensors, medical data, music, art, smell, and taste. This course covers the principles of AI (focusing on deep learning and foundation models), how we can apply AI to novel real-world data modalities, and multimodal AI that can process many modalities at once, such as connecting language and multimedia, music and art, sensing and actuation, and more.