Jingcheng Liang

Hi there! I am a senior undergraduate student in Computer Science at University of Minnesota, Twin Cities. Previously, I studied Economics at South China Normal University.

I am currently a research intern in Minnesota NLP Group, advised by Prof. Dongyeop Kang. I am broadly interested in AI alignment, human–AI interaction, and multi-agent collaboration, with a focus on building reliable, collaborative, and adaptive language technologies.

Human → LM: How can we understand, evaluate, and oversee language models as their capabilities outpace our ability to reliably assess them?
LM → Human: How can language models adapt to diverse human intentions, capabilities, and values to become better collaborators?
LM → LM: What new behaviors emerge when language models collaborate, and how can we design interactions that shape these collective dynamics?

Previously, I conducted research in computational biology advised by Prof. Chad Myers and Dr. Wen Wang. I also participated in the DFRACS project under the supervision of Prof. Tianxi Li.

news

[Apr 06, 2026]	Our papers was accepted to Findings of ACL 2026.
[Mar 13, 2026]	Join Minnesota NLP Group as research intern.

selected publications

Abstain-R1: Calibrated Abstention and Post-Refusal Clarification via Verifiable RL

Haotian Zhai^*, Jingcheng Liang^*, and Dongyeop Kang

Findings of ACL, 2026

Abs PDF Model Data

Reinforcement fine-tuning has been shown to improve the reasoning ability of large language models; however, when faced with unanswerable queries, it often incentivizes models to still produce an answer, leading to guessing and hallucinations. Prior approaches either enforce generic abstention (e.g., simply outputting “I don’t know”) or encourage models to ask post-refusal clarification questions, but fail to supervise the quality of the reasoning underlying refusal, resulting in superficial abstention without substantive clarification. To address these limitations, we propose an RLVR-based training framework that explicitly treats unanswerability and the accuracy of post-refusal clarification as joint learning objectives. Using this framework, we train Abstain-R1, a 3B-parameter model that encourages models to maintain normal answering behavior on answerable queries, perform calibrated refusal on unanswerable queries, and generate semantically correct and informative clarifications following refusal. Experimental results show that, despite being a 3B-parameter model, Abstain-R1 achieves performance comparable to Deepseek-R1 in terms of false-unknown rate, refusal calibration, and clarification quality, while maintaining strong performance on answerable queries.

personal

My favorite ice cream spot in Minneapolis is Pumphouse Creamery. I regard Jane Goodall as my role model.