Dreemur

NLP @Yale | BS CS @ UIUC | Research on LLM Agent | AI X Game | 🦈

[Dreemur @CharlieDreemur]

Can AI organizations learn how to operate?
We introduce RELIC, a framework where recurring collaboration failures become governed, executable protocols that persist across work and member turnover.
Agents learn. Can organizations learn too?
🌐 https://t.co/3OIac98h0g

[Hongyi Du @hongyidu_ai]

Can AI organizations learn how to operate?
We introduce RELIC, a framework where recurring collaboration failures become governed, executable protocols that persist across work and member turnover.
Agents learn. Can organizations learn too?
🌐 https://t.co/JcAPonR7Un

[Dreemur @CharlieDreemur]

@lrzneedresearch 我宣布neu是新的四大

[Dreemur @CharlieDreemur]

@akiyamamizukaze @grok 看看半导体

[Dreemur @CharlieDreemur]

@akiyamamizukaze 天才算术教室

[Dreemur @CharlieDreemur]

AI agents shouldn’t just fail🚨, they should explain why!
Meet🚀 AgentDebugX 🚀open-source debugging framework that turns messy agent trajectories into evidence-backed diagnoses, actionable fixes, and validated reruns.🔍
Try it out in our website: https://t.co/Gve4aWbqO0

[Kunlun Zhu @Kunlun_Zhu]

We’re excited to release AgentDebugX, an open-source toolkit for debugging LLM agents. Instead of only replaying traces, it closes the loop: Detect → Attribute → Recover → Rerun. Paper, code, demo & website below 🧵

[Kunlun Zhu @Kunlun_Zhu]

We’re excited to release AgentDebugX, an open-source toolkit for debugging LLM agents. Instead of only replaying traces, it closes the loop: Detect → Attribute → Recover → Rerun. Paper, code, demo & website below 🧵

[Juanxi Tian @JuanxiTian]

🚀WorldFoundry: Unified World Model Inference & Evaluation Infrastructure

We welcome the community to ⭐ star the repository, submit pull requests, open issues, and contribute new models and benchmarks.

https://t.co/gzBLKbcmWp
https://t.co/Qz1WM4P9B5

[Jindong Wang @jindong_wang92]

Selected as #ACL2026 Oral!

[Dreemur @CharlieDreemur]

@jd92wang congrats!

[Dreemur @CharlieDreemur]

@lrzneedresearch 带带

[Dreemur @CharlieDreemur]

Happy to share that I can finally get 100+ Citation before finishing my undergraduate. Hopefully everything works well for neurips and emnlp in the following days, already exhausted. 😢

[Dreemur @CharlieDreemur]

@akiyamamizukaze 加油喵

[Dreemur @CharlieDreemur]

@demisama_ @MSFTResearch congrats! I also want to join msr 😂

[Dreemur @CharlieDreemur]

@p_song1 congrats!

[Dreemur @CharlieDreemur]

@akiyamamizukaze 酷炫

[Dreemur @CharlieDreemur]

@dviolettchan @jiqizhixin xm女装

[Lorenzo Xiao @ COLM 2026 @lrzneedresearch]

We disagree with @DarioAmodei and @rgblong

We think that AI welfare is essentially Frankfurtian Bullshit

https://t.co/1ebJ8CbGOM

[Dreemur @CharlieDreemur]

@p_song1 @LTIatCMU Congrats!

[Dreemur @CharlieDreemur]

@demisama_ do you plan to go, how is this program. I also just receive this

[Jindong Wang @jindong_wang92]

LLM-based data augmentation is everywhere… but is it safe?

In this #ACL2026 paper, we uncover a critical issue: bias inheritance — when LLM-generated data propagates and amplifies bias in downstream models.

🔍 First systematic study across 10 tasks
⚠️ Bias can help accuracy but worsen fairness
🧠 Identifies 3 key misalignments
🛠️ Proposes token / mask / loss mitigation
This is just the beginning for safer data augmentation.

📄 Paper: https://t.co/awgjPb2T6I