Manoj Shakya, PhD

AI Researcher

[Xudong Han @Xudong07452910]

最近如果想系统补一下强化学习，尤其是 LLM post-training，这套 Berkeley CS 185/285 很值得收藏。

Sergey Levine 的 Deep Reinforcement Learning 2026 春季版课程资料已经完整放出来了，昨天课程录像也开始上传到 YouTube。

课程从 imitation learning、Policy Gradient、Actor-Critic、Q-Learning，一直讲到 Model-Based RL、Offline RL 和 Exploration。

我比较感兴趣的是今年还专门加入了「RL with Sequences & LLMs」。

里面会从 RL 的视角重新讲 RLHF、Verifier Reward、GRPO、KL Regularization 这些现在做 LLM post-training 经常看到的东西。

而且不只是听课，Homework 4 直接要求自己实现 GR-REINFORCE 和 GRPO，再真正跑训练实验做对比。

最近 RL、Agent 和 LLM post-training 连得很紧，这套课正好可以把很多零散的概念重新串起来。

课程：
https://t.co/sWyVEo63ep
YouTube 上搜索 RAIL 的「CS 185/285 (Spring 2026)」播放列表就能看到最新课程录像。

[Andrej Karpathy @karpathy]

Don't think of LLMs as entities but as simulators. For example, when exploring a topic, don't ask:

"What do you think about xyz"?

There is no "you". Next time try:

"What would be a good group of people to explore xyz? What would they say?"

The LLM can channel/simulate many perspectives but it hasn't "thought about" xyz for a while and over time and formed its own opinions in the way we're used to. If you force it via the use of "you", it will give you something by adopting a personality embedding vector implied by the statistics of its finetuning data and then simulate that. It's fine to do, but there is a lot less mystique to it than I find people naively attribute to "asking an AI".

[Manoj Shakya, PhD @shakyamanoj]

PC: Rujal Shakya

[Manoj Shakya, PhD @shakyamanoj]

My seminar talk on AI in action at NAST.

[Manoj Shakya, PhD @shakyamanoj]

Panel Discussion on AI at KCC Conference

[Manoj Shakya, PhD @shakyamanoj]

आज पनि यो मानसिकता हाम्रो समाजमा जिउदै होला…

[Manoj Shakya, PhD @shakyamanoj]

@KarkeeManoj @WazzuCpaas Congratulations

[Manoj Shakya, PhD @shakyamanoj]

@BisheshKhanal7 @NeurIPSConf Congratulations

[Manoj Shakya, PhD @shakyamanoj]

What a great gesture from my friend Tianming! Thank you!
#help
#respect
#gratefulness

[Manoj Shakya, PhD @shakyamanoj]

Github is having issues when creating a repository. Push does not work either. Anyone facing the same problem?
#github

[Manoj Shakya, PhD @shakyamanoj]

https://t.co/3229w0gFel

[Manoj Shakya, PhD @shakyamanoj]

The real story behind Nepal’s wildlife conservation success | 101 East https://t.co/O86lp4EwYJ via @YouTube

[Xavier Bresson @xbresson]

End of the workshop "Artificial Intelligence and Discrete Optimization"
https://t.co/PQEW4OnvyC

Thanks again to @ipam_ucla, the speakers and the participants for an exceptional workshop!

All talks will be available at
https://t.co/EH2BuKVUgC

[yobibyte @y0b1byte]

My take on why its hard to move RL to the real world and why the expectations inflated by stunning results on toy tasks are harmful:

[elvis @omarsar0]

A Categorical Archive of ChatGPT Failures

Comprehensive analysis of ChatGPT failures for categories like reasoning, factual errors, maths, and coding.

If you are developing with LLMs it's important to know these failures. Good to see them documented.

https://t.co/lSrIrfuWcz

[Manoj Shakya, PhD @shakyamanoj]

Are we really this low?
Source: https://t.co/HCPGaZpwRO.
#IQ_Nepal

[Richard Sutton @RichardSSutton]

It is sad to lose the DeepMind office in Edmonton to the Tech layoffs and looming recession. But AI is not going away, and I am more focused than ever on the Alberta Plan for AI research.
https://t.co/DQBl7vmGaO