Shang Yang

Ph.D. Student @MITEECS. Undergrad @Tsinghua_Uni.

[Shang Yang @Shang_mit]

🥳 Thrilled to see our work on TLT featured on the MIT homepage (https://t.co/KHXXqIHwW6) and the cover of MIT News today! 🏛️✨

🚀 2x faster RL training without losing accuracy.

News: https://t.co/f2XkmTLoWT
Paper: https://t.co/x92mPDzK2P
Code: https://t.co/h1zeBsb3lG

[Shang Yang @Shang_mit]

🚀 Introducing TLT (Taming the Long-Tail), an efficient, lossless system that boosts reasoning RL training by mitigating the rollout bottleneck!
🏆 Accepted by #ASPLOS2026

✨ What’s new?
⚡ Enjoy 2× faster end-to-end reasoning RL training
🔒 Lossless on-policy RL — training quality preserved theoretically and empirically
🎁 Get a free, high-quality draft model for efficient deployment

🔗 Github: https://t.co/zVmjaq9CFO
🔗 Paper: https://t.co/RWtX7WYoLv

👇 More in the thread (1/7)

#ASPLOS #EfficientAI #OpenSource #LLMs #Reasoning #ReinforcementLearning

[Shang Yang @Shang_mit]

Come check out our work at CVPR! Excited to chat!

[Zhuoyang Zhang @zhuoyang_zhang]

We‘ll be presenting ForeAct (🌟Highlight) at CVPR 2026:
📍 Poster: Sunday, 3:30PM, ExHall A #95
📍 NVIDIA Tech Talk: Friday, 12:40PM, Booth #211 @NVIDIAAI
Feel free to stop by and chat! Also find our coffee making demo empowered by ForeAct! (https://t.co/a58GwCbgR0)

I’ll be at #CVPR2026 from Jun 4-7. Open to chat about VLA, world models and efficient visual generation!

[rita @rita17371931]

Expert traces can guide RLVR, but full traces may leak key values, executable code, or answer entities — creating an unintended reward-hacking channel.

We propose SMEPO: Semantic Masked Expert Policy Optimization.

Paper: https://t.co/55uoo9rGTw
Code: https://t.co/AH8BiTdpPc

[Shang Yang @Shang_mit]

@YichuanM Congrats!!!

[Han Cai @hancai_hm]

On-policy distillation (OPD) is one of the most effective LLM post-training methods, but it traditionally requires a costly live teacher server throughout training.

In our latest work, Lightning OPD, we show that OPD can be performed fully offline by precomputing teacher log-probabilities before training, reducing OPD to a standard single-model training job.

A key insight is Teacher Consistency: the SFT teacher and OPD teacher must be identical. Otherwise, offline OPD suffers a significant accuracy drop.

Paper: https://t.co/zTW3mm2wWX
Code: https://t.co/7iOB8LezVQ

Contributors: Yecheng Wu, Song Han, Han Cai

[Song Han @songhan_mit]

Honored to join the MIT President’s podcast. We discussed how Efficient AI and model compression techniques can help break the bottlenecks of AI computing: https://t.co/83EIKbf7Dl

[Luke J. Huang @whatthelukh]

We introduce Variance Controlled Policy Optimization (VCPO), a method for explicit variance-targeted controls for policy-gradient objectives in off-policy RL — enabling stable, scalable Async RL training.

✨ Seamlessly integrates into common policy-gradient methods like REINFORCE/RLOO/GRPO 🚀 2.5x faster Async RL training while matching Synchronous RL performance 🧠 Robust training stability under high off-policy settings (at least 128 steps off-policy)
📄Paper: https://t.co/Op29IwenBW
🔗Code: https://t.co/E7PzqmlJX8
🧵👇

[Zhijian Liu @zhijianliu_]

Reasoning LLMs generate very long chains-of-thought, so even small quantization errors add up.

With AWQ, Qwen3-4B drops 71.0 �� 68.2 on MMLU-Pro (~4% relative loss). 😬

ParoQuant fixes this! It keeps only the critical rotation pairs and fuses everything into a single kernel.

Recovers most of the lost reasoning accuracy with minimal overhead — so 4-bit models stay strong at reasoning. 💪💪

[Shang Yang @Shang_mit]

Enhancing your own VLAs with our visual planner. Efficient and easy to integrate! 🥳

📄Paper: https://t.co/aMFjWDv2k1
🔗Code: https://t.co/bzCghtjqu9

#CVPR2026

[Zhuoyang Zhang @zhuoyang_zhang]

We release ForeAct (accepted to CVPR’26🎉), a world model planner powered by visual foresight for VLAs - efficiently, modularly, and at scale.

✨ Seamlessly integrates with VLAs by visual augmentation — no architectural changes required
⚡ Generates high-fidelity 640×480 subgoal images in just 0.33s
🧠 Significantly boosts generalization capability and data efficiency

📄Paper: https://t.co/aOadlsgVqu
🔗Code: https://t.co/aPtnZ0t8Q4

🧵👇

[Hao Kang @GT_HaoKang]

🔥Modifying 2 lines of code and get your agentic serving/rollout up to 3.9x faster losslessly!

⚡️Say hello to ThunderAgent, a fast, simple, and program-aware agentic Inference System.

🥇 We propose a program abstraction to schedule all GPU and CPU resources, the first principled approach for distributed agentic inference and rollout.

🌐 Blog: https://t.co/PAcgTZzlhD
💻 Code: https://t.co/nr7XJj1L7B
📜 Paper: https://t.co/aCD6POzwkU

#AI #ThunderAgent #LLMAgent #Mlsys

1/n

[Shang Yang @Shang_mit]

@Guangxuan_Xiao @MITEECS @thinkymachines congrats @Guangxuan_Xiao

[Zhijian Liu @zhijianliu_]

Holiday cooking finally ready to serve! 🥳

Introducing DFlash — speculative decoding with block diffusion.
🚀 6.2× lossless speedup on Qwen3-8B
⚡ 2.5× faster than EAGLE-3

Diffusion vs AR doesn’t have to be a fight.
At today’s stage:
• dLLMs = fast, highly parallel, but lossy
• AR LLMs = accurate, sequential, but slow

DFlash = diffusion drafts, AR verifies.

[Jack Cook @jackcookjack]

Training LLMs with NVFP4 is hard because FP4 has so few values that I can fit them all in this post: ±{0, 0.5, 1, 1.5, 2, 3, 4, 6}. But what if I told you that reducing this range even further could actually unlock better training + quantization performance?

Introducing Four Over Six, a new method for improving the accuracy of NVFP4 quantization with Adaptive Block Scaling. 🧵

[Shang Yang @Shang_mit]

@Cameron_Chann Really appreciate it!

[Shang Yang @Shang_mit]

🚀 Introducing TLT (Taming the Long-Tail), an efficient, lossless system that boosts reasoning RL training by mitigating the rollout bottleneck!
🏆 Accepted by #ASPLOS2026

✨ What’s new?
⚡ Enjoy 2× faster end-to-end reasoning RL training
🔒 Lossless on-policy RL — training quality preserved theoretically and empirically
🎁 Get a free, high-quality draft model for efficient deployment

🔗 Github: https://t.co/zVmjaq9CFO
🔗 Paper: https://t.co/RWtX7WYoLv

👇 More in the thread (1/7)

#ASPLOS #EfficientAI #OpenSource #LLMs #Reasoning #ReinforcementLearning

[Shang Yang @Shang_mit]

Thank you, Changyu. The blog by Jiajun and Chenyang is excellent!
The idea of applying speculative decoding to RL is gaining traction in the community, and our works approach the problem from complementary perspectives. For example, in TLT, we update the draft model by leveraging rollout bubbles, and the blog updates the drafter jointly with the target model backwards. Additionally, we also introduce optimizations such as adaptive rollout engine that auto-tunes SD configurations to handle the highly dynamic workloads in reasoning RL for better system efficiency. :)

[Shang Yang @Shang_mit]

TLT delivers prominent end-to-end gains:

🔥 1.7×–2.1× training speedup across models at different scales
🔒 Lossless on-policy training — reward curves closely match the baseline

A fast, reliable upgrade for reasoning LLM RL training!

(7/7)

[Shang Yang @Shang_mit]

Rollout workloads are dynamic. The best SD config may change over time — no single setting works well!

👉 TLT auto-tunes SD configurations, always selecting the optimal strategy with zero extra user burden.

(6/7)