Dapeng Jiang @RandomForestDP

CS Phd & Mechanics BS & Aerospace BEng @Tsinghua_Uni AI4Science, Agent and RL. Always happy to chat with people from different background Joined June 2024

Tweets

11
Followers

32
Following

392
Likes

14

Dapeng Jiang @RandomForestDP

20 hours ago

@jordigraumo Just sent the email :)

0 0 1 2K 0

View Details

Awesome AI Dropout Try to collect some students drop their degree and retry in a different country. Want to encourage more to follow their heart and make hard decisions. jdp22.github.io/ai-research-pa…

0 0 0 177 0

View Details

Dapeng Jiang @RandomForestDP

3 weeks ago

@KomatsuSo68 Yes！

0 0 1 6 0

View Details

Dapeng Jiang @RandomForestDP

3 weeks ago

1/6 Excited to share my first-authored NeurIPS paper, Co-Evolve: A “One-to-Many” LLM Fine-Tuning Paradigm!

3 0 2 226 0

View Details

Dapeng Jiang @RandomForestDP

3 weeks ago

@greatxue_33 Thank you Zhongkai!

0 0 0 8 0

View Details

Dapeng Jiang @RandomForestDP

3 weeks ago

@KomatsuSo68 Good question！Independent fine-tunes optimize individual performance. We give each model more credit for solving examples under-covered by its peers. Furthermore, our experiment results show that evolve together is better than one by one.

1 0 1 12 0

View Details

Dapeng Jiang @RandomForestDP

3 weeks ago

6/6 We’ll open-source the code and more details very soon. Looking forward to sharing more and hearing your thoughts!

0 0 0 15 0

View Details

Dapeng Jiang @RandomForestDP

3 weeks ago

5/6 We also introduce seed-replay checkpoints: descendants share one base checkpoint and use compact update logs, reducing the storage needed to maintain multiple models.