Muhan Lin

CMU Master of Robotics. Interested in human-value aligned RL, reward specification, multi-agent collaboration, LLM context scaling!

[Muhan Lin @MuhanLin424433]

Presenting our work which utilizes LLM to assist in individual contribution determination in multi-agent systems! Welcome to my poster if you’re at NeurIPS🌊!

The paper link https://t.co/N6LtQaucfx

[Muhan Lin @MuhanLin424433]

Hey does anyone know some work applying DPO to general non-language tasks with RL?

[Muhan Lin @MuhanLin424433]

We are presenting our awesome work at Riverfront poster session at EMNLP! Looking forwards to talking with everyone!

[Muhan Lin @MuhanLin424433]

🔍Our key contributions💡

- The scale of this reward function naturally reflects an LLM’s ranking uncertainty.
- The method issues uninformative rewards when the LLM is uncertain, avoiding misleading rewards and enabling efficient training even with small error-prone LLMs.

[Muhan Lin @MuhanLin424433]

🚀 Excited to share our latest work: extending RLAIF to work well with small language models which may produce incorrect rankings! 🚀 Catch our poster session at #EMNLP in Miami next Thursday!

📄 Paper: https://t.co/eGXXKfVE3p
🎥 Presentation: https://t.co/6amDzfvRFb

[Muhan Lin @MuhanLin424433]

We introduce a novel potential-based reward specification method that addresses this challenge with promising results.

[Muhan Lin @MuhanLin424433]

Presenting our work which improves robustness of RL based on noisy LLM feedback today at @RLBRew_2024 ! If you are at RLC and would like to chat, please reach out :)!

The paper link is here 👉https://t.co/my2sRztnQe