oh brother
This guy seems to have left our team after the first-gen Qwen model. Our team also did not have a “head of RL” role at that time.
A public back-and-forth erupted on X after Tianhang Zhu posted about his new role leading LLM research at Fundamental Research Labs while referencing his past Alibaba Qwen work. Team members pushed back, stating he left after the first Qwen model and that no head-of-RL position existed on the team at the time.
oh brother
This guy seems to have left our team after the first-gen Qwen model. Our team also did not have a “head of RL” role at that time.
Zhu departed Qwen around May 2024, months before the September 2024 Qwen2.5 release, and the team has noted the absence of any formal RL leadership title during his tenure.
While Zhu is confirmed as a co-author on the original Qwen report and the Qwen2 report, the precise scope of his RL work and any shorthand titles he used remain points of open disagreement with no independent verification available.
Positive users praise Tianhang Zhu's early PPO work on Qwen as technically impressive, while negative users mock the ex-engineer's disputed RL leadership claims as self-promotional and tiresome.
No Digg Deeper questions have been answered for this story yet.
The plot thickens! I guess Tianhang can boast of his RL contributions very legitimately. PPO back then was a really high-risk, high-effort choice. I apologize.
@ChujieZheng Tianhang is the first guy who successfully training first-gen Qwen with PPO and show descent performance improvement.
@ChujieZheng i was responsible for delivery of the RL ckpt. so i think head of rl is accurate description for people to quickly understand what i do
yes i left after first qwen release.
my linkedin https://www.linkedin.com/in/bobzhu has exact details of what i did for years. welcome if interested

alright little bro, here's some free advice you didn't ask for:
for reasons too long to go into here, you don't get the luxury of the benefit-of-doubt in backdating whatever claims of leadership you want to make. other people might get away with it, but not you. unfair, i know.
so you can either be remembered in this hellsite as the guy-who-hello-worlded-a-viral-intro-and-immediately-got-community-noted as a liar...
or delete this bit of internet mishap, and all will be forgotten soon.
you can certainly also double down on these claims and hope for the best, but i assure you the cons of being remembered as a liar far outweigh the benefits, especially when you don't have anything comparable to show for since then.
you'll get another shot at fame, you're already well-positioned for it, so don't waste your integrity on small things like this.
glhfdd

@TianhangZhuzth @ChujieZheng Responsible for and “head of” r two different things

@TianhangZhuzth @ChujieZheng bro giving himself titles

@Zac_labs @TianhangZhuzth @ChujieZheng Probably a cultural/language issue. Boss is responsible for the post he guards and taking care of / disciplining his men
@teortaxesTex more thickening, poor kid
@TianhangZhuzth I noticed that the Qwen2.5/3 and Qwen2.5-Math technical reports appear on your Google Scholar, but I could not find your name in the author lists. Could you clarify your role in these works?
@suchenzang but consider also
I just reviewed the Qwen&Qwen2 tech reports.Qwen1 definitely used PPO, but I’m curious why Qwen2 switched to DPO instead of sticking with it because Tianhang left? PPO is technique used in GLM 5.2! btw I’d really love to know what he actually worked on for Qwen2.5 and Qwen3. 👀
I expect a lot of middling, striverist Chinese from their leading labs to capitalize on this popular idea that "the best and brightest flee to America! hehe brain drain!" and defraud smug wypipos of their money. No, Wenfeng isn't coming. honest advice, learn mafs yourselves
oh brother

@teortaxesTex he is on the qwen 1 and 2 technical report and left before 2.5 to join yi 01ai. Google Scholar automatically update author profile so it could just be faulty update 🤷♂️. Not to say he's clean though 👀

@TianhangZhuzth @ChujieZheng 经典,还能自封

@teortaxesTex Did the RL bro get a $100 million comp package? 🤣😅

@teortaxesTex he did

@TianhangZhuzth @ChujieZheng 谦虚谨慎很重要,别翻车了,祝福

@teortaxesTex 这种操作确实让人很难绷住

@teortaxesTex 还得是硬核技术派真香啊

@teortaxesTex Lmao🤣

@teortaxesTex 这一波含金量确实拉满了承认大佬的眼光准

@teortaxesTex And there was I getting excited

@suchenzang @TianhangZhuzth @ChujieZheng Sir this is a Wendy’s.