Junlin Wang

CS PhD · Duke University

junlin.wang2[at]duke.edu

About Me

I am a 4th year Computer Science PhD at Duke University, advised by Bhuwan Dhingra and previously also advised by Sam Wiseman from 2022-2023.

Before Duke, I work closely with Sameer Singh on Machine Learning Interpretations and Natural Language Processing Projects. I also worked as a research intern in Google Deepmind, Together AI, AWS.

My research experience has been primarily on reasoning agents, long horizon tasks, and continual learning.

News

Achieved best public score on ARC-AGI-3! [tweet]. Blog post and code comming soon!
I will be presenting at NeurIPS 2025!
Summer 2025: Research internship at Google DeepMind.
I will be presenting at EMNLP 2024!

Publications

DSGym: A Holistic Framework for Evaluating and Training Data Science Agents

Fan Nie, Junlin Wang, Harper Hua, Federico Bianchi, Yongchan Kwon, Zhenting Qi, Owen Queen, Shang Zhu, James Zou
arXiv 2026

paper code blog tweet

BibTeX

@article{nie2026dsgym,
  title   = {DSGym: A Holistic Framework for Evaluating and Training Data Science Agents},
  author  = {Fan Nie and Junlin Wang and Harper Hua and Federico Bianchi and Yongchan Kwon and Zhenting Qi and Owen Queen and Shang Zhu and James Zou},
  year    = {2026},
  journal = {arXiv preprint arXiv: 2601.16344}
}

When Does Divide and Conquer Work for Long Context LLM? A Noise Decomposition Framework

Zhen Xu, Shang Zhu, Jue Wang, Junlin Wang, Ben Athiwaratkun, Chi Wang, James Zou, Ce Zhang
ICLR 2026

paper

BibTeX

@inproceedings{xu2026when,
  title     = {When Does Divide and Conquer Work for Long Context {LLM}? A Noise Decomposition Framework},
  author    = {Zach Xu and Shang Zhu and Jue WANG and Junlin Wang and Ben Athiwaratkun and Chi Wang and James Zou and Ce Zhang},
  booktitle = {The Fourteenth International Conference on Learning Representations},
  year      = {2026},
  url       = {https://openreview.net/forum?id=ddQFUuHDDt}
}

Improving Model Alignment Through Collective Intelligence of Open-Source LLMs

Junlin Wang, Roy Xie, Shang Zhu, Jue Wang, Ben Athiwaratkun, Bhuwan Dhingra, Shuaiwen Leon Song, Ce Zhang, James Zou
ICML 2025

paper blog tweet

BibTeX

@inproceedings{wang2025improving,
  title     = {Improving Model Alignment Through Collective Intelligence of Open-Source Models},
  author    = {Junlin Wang and Roy Xie and Shang Zhu and Jue WANG and Ben Athiwaratkun and Bhuwan Dhingra and Shuaiwen Leon Song and Ce Zhang and James Zou},
  booktitle = {Forty-second International Conference on Machine Learning},
  year      = {2025},
  url       = {https://openreview.net/forum?id=K4N9UvsuNB}
}

Think Deep, Think Fast: Investigating Efficiency of Verifier-free Inference-time-scaling Methods

Junlin Wang, Shang Zhu, Jon Saad-Falcon, Ben Athiwaratkun, Qingyang Wu, Jue Wang, Shuaiwen Leon Song, Ce Zhang, Bhuwan Dhingra, James Zou
arXiv 2025

paper tweet

BibTeX

@article{wang2025thinkdeep,
  title   = {Think Deep, Think Fast: Investigating Efficiency of Verifier-free Inference-time-scaling Methods},
  author  = {Junlin Wang and Shang Zhu and Jon Saad-Falcon and Ben Athiwaratkun and Qingyang Wu and Jue Wang and Shuaiwen Leon Song and Ce Zhang and Bhuwan Dhingra and James Zou},
  year    = {2025},
  journal = {arXiv preprint arXiv: 2504.14047}
}

How Much Backtracking is Enough? Exploring the Interplay of SFT and RL in Enhancing LLM Reasoning

Hongyi Cai, Junlin Wang, Xiaoyin Chen, Bhuwan Dhingra
arXiv 2025

paper tweet

BibTeX

@article{cai2025backtracking,
  title   = {How Much Backtracking is Enough? Exploring the Interplay of SFT and RL in Enhancing LLM Reasoning},
  author  = {Hongyi Cai and Junlin Wang and Xiaoyin Chen and Bhuwan Dhingra},
  year    = {2025},
  journal = {arXiv preprint arXiv: 2505.24273}
}

Knowing When to Stop: Efficient Context Processing via Latent Sufficiency Signals

Roy Xie, Junlin Wang, Paul Rosu, Chunyuan Deng, Bolun Sun, Zihao Lin, Bhuwan Dhingra
NeurIPS 2025

paper code project page tweet

BibTeX

@article{xie2025when,
  title   = {Knowing When to Stop: Efficient Context Processing via Latent Sufficiency Signals},
  author  = {Xie, Roy and Wang, Junlin and Rosu, Paul and Deng, Chunyuan and Sun, Bolun and Lin, Zihao and Dhingra, Bhuwan},
  journal = {Advances in Neural Information Processing Systems},
  year    = {2025}
}

Mixture-of-Agents Enhances Large Language Model Capabilities

Junlin Wang, Jue Wang, Ben Athiwaratkun, Ce Zhang, James Zou
ICLR 2025

paper code blog tweet

BibTeX

@inproceedings{wang2025mixtureofagents,
  title     = {Mixture-of-Agents Enhances Large Language Model Capabilities},
  author    = {Junlin Wang and Jue Wang and Ben Athiwaratkun and Ce Zhang and James Zou},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year      = {2025},
  url       = {https://arxiv.org/abs/2406.04692},
}

Reasoning in Token Economies: Budget-Aware Evaluation of LLM Reasoning Strategies

Junlin Wang, Siddhartha Jain, Dejiao Zhang, Baishakhi Ray, Varun Kumar, Ben Athiwaratkun
EMNLP 2024

paper

BibTeX

@article{wang2024reasoning,
  title   = {Reasoning in Token Economies: Budget-Aware Evaluation of LLM Reasoning Strategies},
  author  = {Junlin Wang and Siddhartha Jain and Dejiao Zhang and Baishakhi Ray and Varun Kumar and Ben Athiwaratkun},
  year    = {2024},
  journal = {arXiv preprint arXiv: 2406.06461}
}

ReCaLL: Membership Inference via Relative Conditional Log-Likelihoods

Roy Xie, Junlin Wang, Ruomin Huang, Minxing Zhang, Rong Ge, Jian Pei, Neil Zhenqiang Gong, Bhuwan Dhingra
EMNLP 2024

paper code blog

BibTeX

@article{xie2024recall,
  title   = {ReCaLL: Membership Inference via Relative Conditional Log-Likelihoods},
  author  = {Roy Xie and Junlin Wang and Ruomin Huang and Minxing Zhang and Rong Ge and Jian Pei and Neil Zhenqiang Gong and Bhuwan Dhingra},
  year    = {2024},
  journal = {arXiv preprint arXiv: 2406.15968}
}

Raccoon: Prompt Extraction Benchmark of LLM-Integrated Applications

Junlin Wang*, Tianyi Yang*, Roy Xie, Bhuwan Dhingra
ACL 2024 Findings

paper code

BibTeX

@article{wang2024raccoon,
  title   = {Raccoon: Prompt Extraction Benchmark of LLM-Integrated Applications},
  author  = {Junlin Wang and Tianyi Yang and Roy Xie and Bhuwan Dhingra},
  year    = {2024},
  journal = {arXiv preprint arXiv: 2406.06737}
}

LLM-Resistant Math Word Problem thumbnail

LLM-Resistant Math Word Problem Generation via Adversarial Attacks

Roy Xie, Chengxuan Huang, Junlin Wang, Bhuwan Dhingra
EMNLP 2024 Findings

paper code

BibTeX

@inproceedings{Xie2024adversarial,
  title  = {Adversarial Math Word Problem Generation},
  author = {Roy Xie and Chengxuan Huang and Junlin Wang and Bhuwan Dhingra},
  year   = {2024},
  url    = {https://openreview.net/forum?id=bJz5uGzEe6&referrer=%5BAuthor%20Console%5D(%2Fgroup%3Fid%3Daclweb.org%2FACL%2FARR%2F2024%2FJune%2FAuthors%23your-submissions)}
}

NeuroComparatives: Neuro-Symbolic Distillation of Comparative Knowledge

Phillip Howard*, Junlin Wang*, Vasudev Lal, Gadi Singer, Yejin Choi, Swabha Swayamdipta
NAACL 2024 Findings

paper

BibTeX

@misc{howard2023neurocomparatives,
  title={NeuroComparatives: Neuro-Symbolic Distillation of Comparative Knowledge},
  author={Phillip Howard and Junlin Wang and Vasudev Lal and Gadi Singer and Yejin Choi and Swabha Swayamdipta},
  year={2023},
  eprint={2305.04978},
  archivePrefix={arXiv},
  primaryClass={cs.CL}
}

Maestro: A Gamified Platform for Teaching AI Robustness

Margarita Geleta, Jiacen Xu, Manikanta Loya, Junlin Wang, Sameer Singh, Zhou Li and Sergio Gago Masague
EAAI 2023

paper

BibTeX

@inproceedings{DBLP:conf/aaai/GeletaXLW00M23,
  author    = {Margarita Geleta and Jiacen Xu and Manikanta Loya and Junlin Wang and Sameer Singh and Zhou Li and Sergio Gago Masagu{'{e}}},
  editor    = {Brian Williams and Yiling Chen and Jennifer Neville},
  title     = {Maestro: {A} Gamified Platform for Teaching {AI} Robustness},
  booktitle = {Thirty-Seventh {AAAI} Conference on Artificial Intelligence, {AAAI} 2023, Thirty-Fifth Conference on Innovative Applications of Artificial Intelligence, {IAAI} 2023, Thirteenth Symposium on Educational Advances in Artificial Intelligence, {EAAI} 2023, Washington, DC, USA, February 7-14, 2023},
  pages     = {15816-15824},
  publisher = {{AAAI} Press},
  year      = {2023},
  url       = {https://doi.org/10.1609/aaai.v37i13.26878},
  doi       = {10.1609/AAAI.V37I13.26878}
}

Gradient-based Analysis of NLP Models is Manipulable

Junlin Wang*, Jens Tuyls*, Eric Wallace and Sameer Singh
EMNLP 2020 Findings

paper blog

BibTeX

@inproceedings{wang2020gradientbased,
  Author = {Junlin Wang, Jens Tuyls, Eric Wallace and Sameer Singh},
  Booktitle = {Empirical Methods in Natural Language Processing Findings},
  Year = {2020},
  Title = {Gradient-based Analysis of NLP Models is Manipulable}
}

AllenNLP Interpret: A Framework for Explaining Predictions of NLP Models

Eric Wallace, Jens Tuyls, Junlin Wang, Sanjay Subramanian, Matt Gardner, and Sameer Singh
Demo at EMNLP 2019 Best Demo Award

paper blog

BibTeX

@inproceedings{Wallace2019AllenNLP,
  Author = {Eric Wallace and Jens Tuyls and Junlin Wang and Sanjay Subramanian and Matt Gardner and Sameer Singh},
  Booktitle = {Empirical Methods in Natural Language Processing},
  Year = {2019},
  Title = {AllenNLP Interpret: A Framework for Explaining Predictions of NLP Models}
}

Projects

Link to some of my for fun projects