Menu

Research

Publications by our team.

Publications by team members, written at their universities and labs before or alongside Duffi AI.

  1. EMNLP 2026, main conference

    One Instruction Is a Benchmark: End-to-End Instruction-Following Evaluation with FlexBench

    Yuchen Ge, Yizhao Sun, Wei He, Jinpeng Ou, et al. (13 authors)

    Turns one instruction into its own multi-turn test suite; model rankings and failure modes change from one instruction to the next.

  2. Findings of EMNLP 2026

    Learning When to Think in Latent Visual Space: Reasoning Mode Adaptive Reinforcement Learning

    Paul Jian, Jinpeng Ou, Leo Wei, Qingpei Guo, Jianing Li

    Reinforcement learning that teaches multimodal models when to reason in latent visual space and when in text.

  3. EMNLP 2026, main conference

    To Answer or Not to Answer? Reliability-Oriented Reinforcement Learning for Visual Language Models

    Paul Jian, Julian Wu, Leo Wei, Dylan Chen, Garvin Guo, Qingpei Guo, Jianing Li

    A rubric-based reward that teaches vision-language models to abstain on questions they cannot answer, measured on a new benchmark, Reliable-V-Bench.

  4. Findings of EMNLP 2026

    CAMFT: Conflict-Aware Mergeable Fine-Tuning for Large Language Models

    Jingang Zhou, Haiyang Guo, Yuan Ma, Han Zhu, Xu-Yao Zhang

    Fine-tuning that makes task models easier to merge: each task updates a sparse set of parameters with little conflict across tasks, and merged multi-task models beat those built from standard fine-tuning.

  5. ICML 2026

    From Absolute to Relative: Rethinking Reward Shaping in Group-Based Reinforcement Learning

    Wenzhe Niu, Wei He, Zongxia Xie, Jinpeng Ou, Huichuan Fan, Yuchen Ge, et al. (13 authors)

    Relative rankings in place of absolute scores as the reward in group-based reinforcement learning.

  6. ICASSP 2026

    Think-Augmented Function Calling: Improving LLM Parameter Accuracy Through Embedded Reasoning

    Lei Wei, Xiao Peng, Jinpeng Ou, Bin Wang

    A reasoning parameter in each function signature makes the model think before it fills an argument, with no change to the model and full API compatibility.

  7. ICML 2025

    Asymmetric Decision-Making in Online Knowledge Distillation: Unifying Consensus and Divergence

    Zhaowei Chen, Borui Zhao, Yuchen Ge, Yuhao Chen, Renjie Song, Jiajun Liang

    Online distillation in which students follow what teachers agree on, and teachers keep what sets them apart.

  8. EMNLP 2024, main conference

    Large Language Models Know What is Key Visual Entity: An LLM-assisted Multimodal Retrieval for VQA

    Pu Jian (Paul Jian), Donglei Yu, Jiajun Zhang

    A language model picks out the visual entities a question needs, so retrieval ignores the rest of the image.